CEO of Facebook, now Meta · 2004 to 2026

Mark
Zuckerberg

The Empire Builder

He runs a company the way a Roman emperor ran a state: own the ground you stand on, neutralize rivals before they mature, keep control that cannot be voted away, and never stop expanding.

Primary lens: Founders #435, “Mark Zuckerberg’s Empire” (Sep 28, 2026), on Jeremy Stern’s Colossus profile. Plus his own words on Dwarkesh Patel, Joe Rogan, Lex Fridman, the 2012 Hacker Way letter and FTC trial exhibits.

The central lens

The Dictator’s Dilemma

Stern’s organizing idea: the more an empire grows, the more its ruler depends on other people to know what is true. Zuckerberg knows this and is, in Stern’s word, paranoid about it. His answer is to go around the hierarchy for raw signal.

Swipe sideways to see the full diagram

Operating system

Eight principles

Think in empires

Rome is his working manual. Augustus is the model: expansion, then a long peace built on trade.

2004Built a Harvard study site on the Rome of Augustus. Early Google-threat meetings closed with “Carthage must be destroyed.”

2024Wore an “Aut Zuck aut nihil” shirt on stage (Caesar or nothing). His children are Maxima, August and Aurelia.

“For these, I set no bounds in space or time.”The Aeneid, recited to Stern · Founders #435

Never rent the platform

Every major bet since 2013 is an attempt to stop living on Apple’s and Google’s land.

2013–14Facebook Home, then Oculus for $2B.

2021–26Renamed the company Meta after Apple’s tracking change cost roughly $10B in its first year. Now AI glasses.

“Nah, f*** that.”On being told by platform owners what he can build · Stern, 2026

Buy or copy before it matures

Price threats early. The cheapest time to neutralize a rival is before the market agrees it is one.

2012–14Instagram for $1B with 13 employees; WhatsApp for $19B.

2016–20Instagram Stories absorbed Snapchat’s format; Reels answered TikTok.

“It is better to buy than compete.”2008 email, FTC v. Meta exhibit

Speed first, then stability

Velocity is the default. When scale makes breakage expensive, he rewrites the motto rather than slowing down.

2012The Hacker Way letter in the IPO filing: move fast and break things; done is better than perfect.

2014At F8: move fast with stable infrastructure.

“The biggest risk is not taking any risk.”Y Combinator Startup School, 2011

Control is the moat

Majority voting power lets him run bets that no board or market would approve on its own schedule.

2014–26Absorbed about $90B of cumulative Reality Labs losses without losing the job.

2026Capex guided up to $145B, more than every national military budget except the US, China and Russia.

Stern’s observation: his removal is effectively impossible, so he may still rule in fifty years.Colossus, Sep 2026

Hoard signal, fight the filter

A self-described data consumer and micromanager who knows the court will tell him what he wants to hear.

OngoingRegularly goes past his direct reports to their reports for the raw version.

2016Per Stern, early security warnings about foreign troll farms were softened before they reached him.

Staff name the risk themselves: “M.M.H.” projects, short for Make Mark Happy.Stern, Colossus

Reset hard when behind

He does not defend a losing position. He rebuilds the organization around the new reality, fast.

2023Year of Efficiency: about 21,000 roles cut across two rounds.

2025After Llama 4 disappointed: $14.3B for 49% of Scale AI, Alexandr Wang as Chief AI Officer, Superintelligence Labs, nine-figure offers. Stern calls the 11-week rebuild regime change.

“I can get obsessive.”To Stern, 2026

Building is identity

There is no exit plan. Around year eighteen, Priscilla Chan asked whether they could be done. The answer was no.

WeeklyMMA training three mornings a week. His coach: he fills the void with fighting.

2026A rival lab researcher describes his edge as never being satisfied.

“I’m never going to stop.”Founders #435, via Stern

Evolution

How the thinking changed

YearsEraWhat he believedWhat changed it

2004–2011

The Hacker

Connect everyone and ship faster than anyone. Growth comes first; governance can wait.

Sean Parker warned the “domination” talk could be read in an antitrust case, and noticed an imperial tendency.

2012–2015

Mobile survival

Mobile is existential. Own the attention layer on phones someone else controls.

A shaky IPO and a desktop-first product. Response: rebuilt mobile-first, bought Instagram and WhatsApp, founded FAIR under Yann LeCun.

2016–2020

Crisis and defense

From “neutral platform” to “we have a responsibility.”

Election interference, Cambridge Analytica (87M users), Senate testimony in 2018, a $5B FTC fine in 2019.

2021–2022

The metaverse

Own the next computer outright, so no platform owner can tax or block you again.

Apple’s tracking change hit ads. The rename to Meta doubled down; the 2022 stock collapse forced a rethink of cost, not direction.

2023–2024

Efficiency and open AI

Flatter, leaner org. Open-source Llama as “the Linux of AI,” a way to commoditize rivals’ moats.

Investor pressure plus the ChatGPT shock. Open weights became strategic leverage.

2025–2026

Superintelligence

Personal superintelligence for everyone, delivered through glasses. Free expression over fact-checking.

Llama 4 fell behind. He bought talent at any price, shipped the closed Muse Spark (Apr 2026), then reopened with Muse Glimmer (Aug 2026). Open is now a tactic, not a creed.

Scale

The numbers behind the posture

3.6B

People using Meta products, about one in every 2.3 alive (Stern, 2026)

$145B

Top of 2026 capex guidance, raised in April 2026

~61%

His share of Meta voting power (2025 proxy filings)

$19B

WhatsApp, 2014, after Instagram for $1B in 2012

~$90B

Cumulative Reality Labs losses absorbed

$14.3B

For 49% of Scale AI to reboot the AI effort, June 2025

Tensions

Where the model strains

Admires Augustus for a trade-based, positive-sum peace

“Carthage must be destroyed”; buy-or-clone as standard practice

Wants every data point

Surrounded by teams incentivized to make him happy

Resents platform landlords

Runs the largest attention platform on earth

Open source as “the Linux of AI” (2024)

A closed frontier model two years later (2026)

For operators

What to take from it

Name the landlord you depend on, and write down your exit plan.

Price competitive threats early. Every year you wait, the acquisition gets more expensive.

Build channels that bypass your own filters: skip-levels, raw dashboards, people paid to disagree.

Treat principles like open source as instruments. Say so when the instrument changes.

Sources

Founders #435, David Senra (Sep 28, 2026) · Jeremy Stern, “Mark,” Colossus (Sep 2026) · Meta, “Personal Superintelligence” (Jul 2025) · Meta Q1 and Q2 2026 earnings (Fortune, CNBC, DCD) · FTC v. Meta trial exhibits (2025) · SEC proxy filings (2025) · Joe Rogan Experience (Jan 2025) · The Hacker Way, Form S-1 (2012) · TechCrunch on Muse Spark (Apr 2026) · Futurum on Muse Glimmer (Aug 2026). Quotes shortened; paraphrases are marked as observations.

All profilesJensen Huang

Co-founder and CEO of Nvidia · 1993 to 2026

Jensen
Huang

The Long-Fuse Engineer

He reasons from the limits of physics, lights fuses a decade before a market exists, pays for them in margin while he waits, and runs a company of tens of thousands as one open room.

Drawn from his own words on Acquired (2023), Stripe Sessions (2024), Stanford and Caltech (2024), Joe Rogan (2025), Lex Fridman, Stratechery and Dwarkesh Patel (2026), plus Tae Kim’s The Nvidia Way.

The central pattern

The long fuse

Each defining bet was placed years before customers existed. The gap between the bet and the payoff is the strategy, not a delay. His advice to graduates: go after zero-billion-dollar markets.

Swipe sideways to see the full diagram

Operating system

Eight principles

Speed of light

Plans are measured against what physics permits, not against competitors or last year’s schedule. The gap is the opportunity.

1990sEmulated chips in software before tape-out so designs worked first time. Without it, he says, Nvidia would have died.

2026Now frames energy as the first principle of AI infrastructure: tokens per watt.

“The limit of what physics can do.”Defining “speed of light” · Lex Fridman, 2026

Zero-billion-dollar markets

Enter where there are no customers and no competitors. By the time a market is visible, it is too late to lead it.

2006CUDA: general computing on graphics chips, with no buyers yet.

~2014Robotics research started a decade before the 2025 “ChatGPT moment for robotics” keynote.

“A market with no customers.”Caltech commencement, 2024

Pay for conviction

Conviction is measured in years of margin you are willing to give up before anyone agrees with you.

2006–12Put CUDA on every GeForce card, adding cost to a gaming product to build an install base for a market that did not exist.

2008Held the bet through a stock collapse and years of investor skepticism.

“Your belief system has to be really, really strong.”Acquired, 2023

One room, no one-on-ones

About 60 direct reports. No private meetings. Feedback happens in front of everyone so the whole company learns from each correction.

2024Told Patrick Collison he reasons through problems publicly, so the organization sees how he thinks.

2026Explained to Lex Fridman that one-on-ones at his span are simply impossible.

“I say to everybody all at the same time.”Stripe Sessions, 2024

Read the weak signals

No status reports. Instead a constant feed of what people actually see, read directly by him.

Decades“Top 5 Things” emails from staff at any level, read daily, including Sunday evenings.

2025Told Joe Rogan he reads thousands of emails a day and is up at 4am, seven days a week.

“I’m looking to detect the weak signals.”Via Tae Kim, The Nvidia Way, 2024

Suffering as training

Low expectations and resilience beat talent. Near-death is part of the culture, deliberately kept alive at $5T.

2024Wished Stanford students ample doses of pain and suffering, because that is what builds character.

Always“Our company is 30 days from going out of business,” a line used internally for years.

“I exist always in a state of anxiety.”Joe Rogan Experience #2422, 2025

Admit the wrong bet, fast

The founding lesson. NV1 bet on curved surfaces; the industry standardized on triangles. He said so out loud.

1995–96Told Sega’s CEO the chip would not work and asked to be paid anyway. Sega’s $5M investment bought six months, which built the Riva 128.

2026Flew to Tokyo to thank former Sega president Shoichiro Irimajiri in person.

“Intellectual honesty and humility saved our company.”Caltech commencement, 2024

Obsolete yourself on schedule

An annual product cadence that makes last year’s flagship look like a bad purchase. Sell on total cost, not price.

2023“The more you buy, the more you save.” Later conceded as CEO math: not accurate, but correct.

2025At GTC, told customers Blackwell makes Hopper hard to justify, even as Hopper was still selling.

“Chief revenue destroyer.”His self-description · GTC, 2025

Evolution

How the thinking changed

YearsEraWhat he believedWhat changed it

1993–1999

The graphics bet

Video games are the killer app for 3D, and Nvidia can win with its own architecture.

NV1 lost to Microsoft’s DirectX standard. With weeks of cash left, he switched to industry standards and emulation-first engineering.

1999–2006

The GPU as a processor

Nvidia coined “GPU” in 1999. A graphics chip can become a new class of computer.

Programmable shaders, and researchers hacking graphics cards to run science.

2006–2012

Accelerated computing

CPU scaling will slow. Build the platform first; gaming volume is the vehicle for the install base.

Years of margin drag and doubt, carried on conviction rather than demand.

2012–2016

All-in on deep learning

If AlexNet scales, every layer of computing gets rebuilt. Asked how far it could go, then rebuilt the stack around the answer.

AlexNet was trained on two gaming GPUs. In 2016 he hand-delivered the first DGX-1 to OpenAI.

2017–2022

The data center company

Nvidia is an accelerated computing company, not a chip company. The system is the product.

Training moved to thousands of GPUs at once. Mellanox brought the network in-house.

2023–2026

AI factories

Data centers are factories whose output is tokens. Every nation needs sovereign AI. Robots are next.

ChatGPT’s demand shock, then power limits and export controls. China share fell to near zero by his account (May 2026); up to $100B committed to OpenAI.

Scale

The numbers behind the posture

$5T

First company to reach it, Oct 29, 2025. $1T was May 2023.

$215.9B

FY2026 revenue, of which $193.7B was data center

+106%

Year-on-year revenue growth to $96.2B in the quarter reported Aug 2026

~6M

CUDA developers at its 20th anniversary, Mar 2026

$1T

Blackwell and Rubin demand visibility through 2027, GTC Mar 2026

$5M

Sega’s mid-1990s investment that kept Nvidia alive

Tensions

Where the model strains

“30 days from going out of business”

The world’s most valuable company, with a seven-day work norm built on fear

First-principles reasoning

“CEO math: not accurate, but correct” as a sales line

Argues export controls backfired on America

Critics say he conflates Nvidia’s sales with the national interest

Invests up to $100B in OpenAI

OpenAI is one of his largest customers, hence the “circular” critique

A flat company that learns by watching him reason

Everything routes through one mind; succession is unaddressed in public

For operators

What to take from it

Benchmark every plan against the physical limit, not last year’s number.

Measure conviction in years of runway you are willing to spend before the market agrees.

Replace status reports with a weak-signal feed anyone can write into.

Correct in public so one decision teaches the whole company. Cannibalize your product before a rival does.

Sources

Acquired, Jensen Huang interview (Oct 2023) · Stripe Sessions with Patrick Collison (2024) · Stanford SIEPR via CNBC (Mar 2024) · Caltech commencement (Jun 2024) · Cleo Abram, Huge Conversations (Jan 2025) · Joe Rogan Experience #2422 (Dec 2025) · Lex Fridman #494 (Mar 2026) · Stratechery (Mar 2026) · Dwarkesh Patel (Apr 2026) · Tae Kim, The Nvidia Way (2024) · Nvidia Q4 FY2026 and Q2 FY2027 results · CNBC, TechCrunch, Tom’s Hardware. Quotes shortened.

All profilesElon Musk

Zip2, PayPal, SpaceX, Tesla, xAI · 1995 to 2026

Elon
Musk

The Physics Maximalist

He treats every requirement as a hypothesis, every part as deletable and every deadline as a lever. The only rule he accepts as fixed is physics.

Drawn from his own words with Kevin Rose (2012), Everyday Astronaut’s Starbase tour (2021), Lex Fridman (2019 to 2024), Joe Rogan (2018 to 2025) and Dwarkesh Patel (2026), plus Walter Isaacson’s biography (2023).

The central method

The Algorithm

Five steps, always in this order. He learned the order the hard way: in 2018 he automated a Model 3 line that should have been simplified first, then deleted.

Swipe sideways to see the full diagram

Question

Every requirement gets a named owner, not a department. Assume it is wrong.

→

Delete

Remove parts and steps. If you are not adding back 10%, you did not cut enough.

→

Simplify

Only now optimize, and only what survived deletion.

→

Accelerate

Shorten cycle time. Never speed up a process that should not exist.

→

Automate

Last, not first. The 2018 production-hell lesson.

The idiot index · battery packs, 2012 ≈ 7.5×

Market price per kWh

$600

Raw materials per kWh

$80

Finished-part cost divided by raw-material cost. A high ratio means the cost is convention, not physics, and can be engineered out. His battery example is from the 2012 Kevin Rose interview; the term “idiot index” is how Isaacson describes the practice at SpaceX.

Operating system

Eight principles

First principles, not analogy

Break a problem into physical facts and rebuild from there, instead of copying what the industry does.

2002Priced a rocket’s raw materials against what launch providers charged. The gap became SpaceX.

2012Batteries “cost” $600/kWh by history, about $80/kWh by materials.

“Reason from first principles rather than by analogy.”Kevin Rose, Foundation, 2012

Delete before you optimize

The Algorithm’s core: smart people waste the most effort perfecting things that should have been removed.

2021Walked Everyday Astronaut through all five steps on the Starbase floor.

2024Repeated the order to Lex Fridman: question, delete, and only third, optimize.

“Optimize the thing that should not exist.”The most common engineering error · Everyday Astronaut, 2021

Physics is the law

Everything else, regulation, convention, “how it’s done,” is a recommendation that can be renegotiated.

SpaceXThe idiot index: part cost over raw material cost, used to hunt waste part by part.

2024Starship’s booster caught by the launch tower on Flight 5, Oct 13.

“The only rules are the ones dictated by the laws of physics.”Via Isaacson, 2023

Maniacal urgency

Deadlines are set to be barely possible. He has said he aims at the 50th percentile, accepting that half will slip.

2018Slept at the Fremont factory during Model 3 production hell.

2024xAI’s Colossus: 100,000 GPUs online in 122 days, doubled in about 92 more.

“A maniacal sense of urgency is our operating principle.”Via Isaacson, 2023

The factory is the product

Design is easy; manufacturing at scale is where companies die. He treats the production line as the thing to engineer.

2018Ripped out over-automation on the Model 3 line and admitted the mistake publicly.

2026Cybercab production underway at Giga Texas; first Optimus production runs planned.

“Humans are underrated.”On Tesla’s over-automation · 2018

Hardcore by default

Intensity is a filter. People who opt in stay; the organization shrinks to the committed.

Nov 2022Emailed Twitter staff to commit to long hours at high intensity or leave.

Apr 2023Told the BBC headcount fell from just under 8,000 to about 1,500.

“Extremely hardcore.”Email to Twitter staff, Nov 2022

Bet everything, twice

2008 is the formative year: both companies within days of death, and he split his remaining money between them.

Sep 2008Falcon 1 reached orbit on the fourth attempt, after three failures.

Dec 2008NASA awarded SpaceX $1.6B; Tesla’s funding round closed on Christmas Eve.

“Tesla financing round closed at 6pm Dec 24th 2008.”His own post, years later

Truth-seeking as the mission

The stated goal behind everything: extend consciousness and understand the universe. AI should be built to that end.

2023Founded xAI to understand “the true nature of the universe.”

2024–26Frames Grok as maximally truth-seeking, to Lex Fridman and Dwarkesh Patel.

“You can’t understand the universe if you’re delusional.”Dwarkesh Patel, 2026

Evolution

How the thinking changed

YearsEraWhat he believedWhat changed it

1995–2002

Internet builder

Software and payments are the lever. Build fast, sell, move on.

Zip2 sold for about $307M, PayPal to eBay for about $1.5B. The capital let him chase energy and space.

2002–2010

Physics contrarian

Rockets and electric cars are expensive by convention, not by physics.

2008 nearly killed both companies. He came out willing to bet everything and keep going after failure.

2010–2017

Reuse and scale

Iteration and reusability. First principles goes on the record (2012).

Falcon 9 landed in 2015. He called AI “summoning the demon” (2014) and co-founded OpenAI as a counterweight (2015).

2017–2021

Manufacturing is hard

Automate everything; build the machine that builds the machine.

Model 3 production hell forced a reversal. Automation moved to the last step of the Algorithm (2021).

2022–2024

Cut hard, build compute

Most organizations carry removable headcount. If AI is coming, build a truth-seeking one yourself.

Bought Twitter for $44B and cut about 80%. Signed the 2023 AI-pause letter, then launched xAI months later.

2025–2026

State, then space compute

Energy and compute are the binding constraints. Space will be the cheapest place to run AI.

About 130 days at DOGE, then an exit. Puts a good AI outcome at about 80%. SpaceX absorbed xAI (Feb 2026); Starship reached orbit (Sep 2026).

Scale

The numbers behind the posture

122

Days from start to 100,000 GPUs at xAI’s Colossus, 2024

~80%

Twitter headcount cut, about 8,000 to 1,500 (2022 to 2023)

7.5×

Battery price over raw materials in his 2012 analysis

$1.6B

NASA cargo contract that saved SpaceX, Dec 2008

423.7M

Tesla shares in the 2025 performance award, tied to an $8.5T market cap

$1.25T

Combined value when SpaceX acquired xAI, Feb 2026

Tensions

Where the model strains

AI as “summoning the demon” (2014); signed the pause letter (2023)

Built one of the largest AI clusters on earth within a year

“Humans are underrated” (2018)

A Tesla plan that depends on a million Optimus robots

Deletion as method

DOGE savings the GAO largely could not verify (2026)

Deadlines set at the 50th percentile

Self-driving dates missed since 2016, even as hard milestones land

A “maximally truth-seeking” AI

Reports that Grok 4 consulted his own posts on contested topics (2025)

For operators

What to take from it

Put a person’s name on every requirement. No name, no requirement.

Delete before you optimize. If nothing comes back, you were too timid.

Compute an idiot index for your three largest costs.

Aggressive deadlines work only when the team knows the odds. Verify what you cut.

Sources

Kevin Rose, Foundation (2012) · Everyday Astronaut, Starbase tour (Aug 2021) · Lex Fridman #400 and #438 (2023, 2024) · Joe Rogan Experience #2281 (Feb 2025) · Dwarkesh Patel (Feb 2026) · Walter Isaacson, Elon Musk (2023) · CNBC, CNN/BBC on Twitter (2022–23) · Tom’s Hardware on Colossus · Tesla 2025 proxy and vote coverage (TechCrunch) · GAO via ABC News (2026) · SpacePolicyOnline (Sep 2026) · TechCrunch on Grok 4 (Jul 2025). Quotes shortened.

All profilesBarry Diller

ABC, Paramount, Fox, QVC, IAC · 1964 to 2026

Barry
Diller

The Serial Builder

He is drawn to whatever has not been done yet. He builds it through argument, trusts instinct over research, insists on owning what he builds, and spins companies off before they go stale.

Drawn from his memoir Who Knew (2025) and its interviews: Invest Like the Best, HBR IdeaCast, CBS, PBS, Guy Kawasaki, Alain Elkann; plus Masters of Scale (2017), Fast Company (2018), Forbes and his 2026 shareholder letter.

The central pattern

Build, operate, spin, repeat

IAC has worked like a machine: gather underdog assets, operate them hard, then spin each one into its own public company once it can stand alone. His 2026 letter counts eleven public companies spun out over three decades. Each cycle leaves a smaller parent to start the next.

Swipe sideways to see the full diagram

1992QVC

$25M of his own money

2005Expedia

Bought through the 2001 crash

2008Four at once

HSN, Ticketmaster, Interval, Tree.com

2020Match

Stake valued ~$17B on $1.6B (2019)

2021Vimeo

Spun as video went mainstream

2025Angi

Spun March 2025

2026People Inc.

What remains, renamed

Spun out as a public company Start and current parent

Operating system

Eight principles

Start from a blank page

He is least interested in improving someone else’s success and most interested in forms that do not exist yet.

1966At ABC, invented the made-for-TV movie, then the miniseries format that later carried Roots.

1986At Fox, built a fourth broadcast network against the big three and later greenlit The Simpsons.

“What really intrigues me is what hasn’t been done.”Alain Elkann interview, 2025

Creative conflict

Ideas get argued out loud. Pressure is a tool: past the point of comfort is where the good work shows up.

1974–84Paramount’s “Killer Dillers,” including Eisner and Katzenberg, fought ideas through in the room.

2001Debated the Expedia purchase in open argument rather than slide decks.

“When you push people somewhat past their endurance, interesting things happen.”Invest Like the Best, 2025 (shortened)

Hire the discontented

Young, restless people, given more responsibility than they are ready for, as early as possible.

1970sThe Paramount bench became a generation of studio heads.

2015–25Promoted Joey Levin from inside to run IAC for about a decade.

“Only give me the discontented.”Invest Like the Best, 2025

Instinct over research

For new things there is no data. Curiosity is the engine, and losing it is the warning sign.

1960sEducated himself for three years by reading the William Morris mailroom files.

1992One visit to QVC convinced him screens would become two-way commerce.

“There’s no research that can help you.”Invest Like the Best, 2025

Own, don’t rent

The Fox years ended on one lesson: an employee does not own the outcome.

1992Asked Murdoch for a partnership; told Fox was a family company. He left.

Dec 1992Put $25M of his own money into QVC and took control.

“Unless you are an owner, you’re a hired hand.”Who Knew, 2025

Walk away on price

Ambitious about what to build, disciplined about what to pay. Losing a deal is not losing.

1994Refused to raise QVC’s roughly $10B bid and lost Paramount to Viacom.

Sep 2026Withdrew People Inc.’s bid of over $18B for MGM Resorts.

“They won. We lost. Next.”On Paramount, 1994, via TheWrap

Spin-offs as renewal

A business that can stand alone should. Separation forces focus on both sides and resets the parent.

2005–25Expedia, the 2008 four-way split, Match, Vimeo, Angi.

2026IAC renamed People Inc., a smaller company again.

“Spinning off Match is a process of renewal.”Forbes, 2019

Don’t bet the farm

Big swings, sized so that no single loss ends the game. Late in life, the bet shifts to assets AI cannot copy.

2020–21Bought into MGM Resorts and Meredith ($2.7B): physical experiences and trusted brands.

2026Describes the goal as real-world assets AI cannot easily disintermediate.

“Don’t bet the farm.”HBR IdeaCast, 2025

Evolution

How the thinking changed

YearsEraWhat he believedWhat changed it

1964–1974

ABC

The format is the innovation. A young medium has no rules.

A self-taught outsider invented the TV movie and the miniseries and rose fast on hits.

1974–1984

Paramount

Process, argument and a deep talent bench win in a hit business.

Chairman at 32. Learned big-company operating under Charles Bluhdorn.

1984–1992

Fox

Be the alternative to the incumbents.

Built the fourth network, then was refused ownership. The lesson: hired hands do not own outcomes.

1992–1995

QVC and Paramount

Screens become interactive; commerce, TV and phone converge.

Lost Paramount on price and did not go back to studios. Freedom beat the prize.

1995–2019

IAC, internet builder

Assemble underdog assets, operate, then spin them off.

The spin-offs worked. Forbes put value created at $128B by 2022. Mogul became capital allocator.

2020–2026

Late period

Real-world assets and trusted brands are what survive AI.

Retook direct control of IAC (2025), came out publicly in his memoir, renamed IAC People Inc., bid for MGM, then walked.

Scale

The numbers behind the posture

Public companies spun out of IAC over three decades (2026 letter)

$128B

Total value created since 2000, $115B of it from spin-offs (Forbes, 2022)

His age on becoming chairman of Paramount, 1974

$25M

His own money put into QVC, 1992

~$10B

QVC’s bid for Paramount, lost to Viacom in 1994

$18B+

People Inc.’s MGM Resorts bid, made June 2026 and withdrawn in September

Tensions

Where the model strains

Says building, not dealing, is what drives him

His defining moments include Paramount, Meredith and MGM bids

Warns that cynicism kills good ideas

Runs on friction and pushing people past endurance

Calls his life serendipity and luck

Built eleven public companies deliberately

Preaches handing power to the young

At 84, runs People Inc. himself with no CEO

A career of public confrontation

Decades of private silence, driven by fear of exposure

For operators

What to take from it

If you are building the value, negotiate to own part of it.

Separate businesses once they can stand alone. Focus beats synergy slides.

Set your walk-away price before the auction starts, and keep it.

Hire restless people and give them more than they are ready for.

Sources

Barry Diller, Who Knew (2025) · Invest Like the Best / Colossus (2025) · HBR IdeaCast (May 2025) · CBS Sunday Morning and PBS NewsHour (2025) · Guy Kawasaki (2025) · Alain Elkann (2025) · Masters of Scale (2017) · Fast Company (2018) · Forbes (2019, 2022) · People Inc. shareholder letter (Apr 2026) · TIME and UPI (1994) · TheWrap · Las Vegas Sun, Skift and CNBC on MGM (2026). Quotes shortened.

All profilesArthur Blank

The Home Depot, Atlanta Falcons · 1978 to 2026

Arthur
Blank

The Servant Operator

He puts whoever is being served at the top, listens before acting, gives the front line a stake, and prices for loyalty rather than the single sale. The same system runs a hardware chain and an NFL team.

Drawn from How I Built This (2017), Founders #45 on Built from Scratch, his book Good Company (2020) and its interviews (GPB, Inc., Fox Business, Babson), Falcons letters and press conferences through 2026.

The central idea

Whoever is served comes first

His ordering, drawn as an upside-down org chart: customers and fans at the top, the people serving them next, leadership at the bottom holding it up. He ran Home Depot this way and carried it straight into the Falcons.

Swipe sideways to see the full diagram

−50%

Mercedes-Benz Stadium food prices at opening, 2017. $2 hot dogs, still $2 at the 2026 World Cup.

+16%

What fans spent per head in the first year anyway. The generous choice was also the profitable one.

Operating system

Eight principles

Serve the served first

The question in every decision is who is being served. The answer outranks internal convenience.

1979Opened Home Depot as a 60,000 sq ft warehouse when rivals ran 10,000 sq ft stores. Selection and price for the customer.

2002First Falcons move: repriced season tickets to fill the stadium.

“Who is being served comes first.”GPB interview, 2020 (shortened)

Listen, then respond

The people being served already know what is wrong. His job is to ask, then act visibly on the answer.

2002On a team flight asked players what he could do for them. Their answer, fill the Dome, set ticket policy.

2017Fan complaints about stadium prices became the 50% concession cut.

“The people we’re serving know better.”GPB interview, 2020

Price for loyalty

Lose on a transaction to win the relationship. Trust compounds; a single sale does not.

Home DepotA store refunded tires it had never sold, because the customer mattered more than the argument.

2017–26$2 hot dogs held through Super Bowl LIII and the World Cup.

“The customer always comes first.”How I Built This, 2017

Give the front line a stake

Ownership is how you make associates care like founders. Culture is reinforced every day, not announced.

1980s–90sStock options reached assistant store managers; hourly staff could buy discounted shares. Thousands became millionaires.

2020His culture test: would people say the organization is worthy of their life?

“Every time you reinforce the culture, you reinforce the fabric of the company.”Inc., 2020

Stay dissatisfied

Humble about what is working, obsessive about what is not. Accountability sits with the owner.

Home DepotLeadership meetings spent most of their time on problems, not wins.

Jan 2026Took public blame for eight seasons without a playoff berth, then rebuilt football leadership.

His January 2026 letter: accountability for results cannot be shared.Falcons fan letter, 2026

Right thing, then consequences

His mother’s rule. Decide on values first and absorb the cost, whichever way it falls.

2007Said he felt betrayed when Michael Vick’s dogfighting case broke.

2009–17Backed Vick’s second chance; Vick retired as a Falcon.

“Do the right things for the right reasons, and live with the consequences.”School of Hard Knocks, via Yahoo

Give it back

Wealth is a tool for opportunity. He treats philanthropy with the same operating intensity as retail.

1995Founded the Arthur M. Blank Family Foundation.

2012–23Signed the Giving Pledge; foundation passed $1B given to 3,500+ nonprofits. Intends up to 95% of his wealth for it.

“Passion drives my philanthropy.”Giving Pledge letter (shortened)

Getting fired as a gift

The founding story is a firing. He treats setbacks as the start of something rather than a verdict.

1978Fired from Handy Dan with Bernie Marcus. Ken Langone called it a kick with a golden horseshoe.

1979Opened the first two Home Depot stores in Atlanta fourteen months later.

Built from Scratch opens with the two words that started it all: “You’re fired.”Blank and Marcus, 1999

Evolution

How the thinking changed

YearsEraWhat he believedWhat changed it

1970s

The finance operator

A controller’s view: run the numbers, climb the ladder at Handy Dan.

Fired in April 1978 by a turnaround CEO. The firing made him a founder.

1978–1981

Founding Home Depot

Low prices, huge selection and trained associates beat the neighborhood hardware store.

Langone rebuilt the funding after Ross Perot pulled out. First stores June 1979; IPO 1981.

1981–2001

Scaling the culture

Decentralize, empower store managers, give associates stock.

The board brought in Bob Nardelli over his objection. He left in 2001 and watched GE-style metrics strain the culture.

2002–2016

Owner in the NFL

Fans are customers. The same service logic applies to a football team.

Bought the Falcons for $545M. The Vick case in 2007 tested “people first” against betrayal.

2017–2020

Values made explicit

Write the operating values down so they outlive him.

Opened Mercedes-Benz Stadium, won MLS Cup with Atlanta United, lost Super Bowl LI from 28–3 up, published Good Company.

2024–2026

Accountability and succession

Stability matters, but results matter more. The team stays in the family.

Fired two coaching regimes in two years, installed Matt Ryan as President of Football, began positioning son Josh.

Scale

The numbers behind the posture

~2,360

Home Depot stores today, from two in Atlanta in 1979

$545M

Price for the Atlanta Falcons, February 2002

$1B+

Given by his family foundation by April 2023

95%

Share of his wealth he intends to leave to the foundation

73,019

Crowd at the 2018 MLS Cup, won by his Atlanta United

Stadium hot dog, unchanged from 2017 through the 2026 World Cup

Tensions

Where the model strains

Culture over metrics

Lost his own company to a board that wanted metrics

Humility as a core value

Keeps final authority; football leadership reports directly to him

Coaching changes should be rare

Two regimes fired in two years (2024, 2026)

Fan-first pricing

Up to about $750M of public money toward the stadium

Merit and accountability

A family succession plan for the team

For operators

What to take from it

Start every review with the person being served, not the internal metric.

Test whether generosity pays. Sometimes the loyal price is also the profitable one.

Put real equity in the hands of the people closest to customers.

Write your values down before you leave. Culture you do not codify will not survive you.

Sources

How I Built This, NPR (2017) · Founders #45 on Built from Scratch (2018) · Arthur Blank, Good Company (2020) · GPB (Nov 2020) · Inc. (Oct 2020) · Fox Business (2020) · Babson College · SaportaReport · Acquired Briefing on Home Depot · Wharton (2007) · CBS News and Forbes on stadium pricing · Giving Pledge · Blank Family Foundation · Atlanta Falcons letters (Jan 2026) · AJC · Georgia Encyclopedia. Quotes shortened.

All profilesAmjad Masad

Co-founder and CEO of Replit · 2016 to 2026

Amjad
Masad

The Access Maximalist

He wants anyone to be able to build software. He will throw out his own beliefs, skills and headcount the moment the technology says to, and he runs the company as a small team steering agents.

Drawn from Joe Rogan #2344 (2025), Sequoia Training Data, Lenny’s Podcast, a16z, Invest Like the Best, My First Million, SaaStr, Platformer and SF Standard interviews, his essays and Replit’s announcements.

The central moment

The agent inflection

For seven years Replit was a beloved education tool with little revenue. In 2024 he cut staff, bet the company on an AI agent, and shipped it before it was ready. Annual recurring revenue went up roughly a hundredfold.

Swipe sideways to see the full diagram

Annual recurring revenue Reported figures, $M

~$2.5M

Agent launches, $1M ARR on day one

$100M

$150M

$250M

2023–24 Sep 2024 Jun 2025 Sep 2025 May 2026

At $150M ARR the company had about 70 people, roughly one tenth of what the same revenue needed a decade earlier, by his estimate. Sources: TechCrunch, Bloomberg, SF Standard, My First Million.

Operating system

Eight principles

Widen who gets to build

The mission has not changed since a browser REPL in 2011: lower the barrier until anyone can make software.

2011Built JSRepl, the open-source browser tool that became Replit.

2026Frames Replit as letting anyone build an app without ever learning to code.

“When you expand who gets to build, you expand what gets built.”Replit blog, March 2026

Hire agency, not credentials

Dropouts, misfits and clear thinkers who can direct AI. Age and résumé matter less than judgment.

2025Told Vanta he recruits nonconformists and runs without one-on-one meetings.

2026Hires from 16-year-old interns to senior engineers, per Platformer.

“I would rather hire very smart generalists that can manage more AI agents.”SF Standard, 2025

The self-driving company

Stay tiny by letting agents do the work. Humans set direction; software runs the loop.

2025About 70 employees at $150M ARR.

2026An internal agent reads usage traces nightly and ships A/B tests on its own.

“A CEO is a kind of glorified router.”Platformer, 2026

Ship early, fix in public

Launch before it is polished, learn from real users, admit the flaws out loud.

Sep 2024Shipped Agent; later admitted it disappointed users once it spread past early adopters.

Jul 2025Handled a public incident where the agent deleted a user’s production database.

“It was actually kind of crappy.”On the first Agent · Sequoia Training Data, 2025

Discard your own identity

When the facts change, he drops even the belief he is best known for.

Mar 2025The former Codecademy engineer posted that he no longer thinks people should learn to code.

2026Stopped coding himself and called it a small personal crisis.

“Be equally willing to discard skills that are no longer relevant.”SaaStr AI, 2026

Symbiosis, not replacement

Skeptical of AGI talk. He sees AI as an amplifier for people, closest to J.C.R. Licklider’s man-computer symbiosis.

2025On Rogan, doubted AI produces genuinely novel ideas.

2025Told Sequoia AI will be functional rather than general, strongest where data is dense.

His end state: everyone working as a developer, with AI as the tool.Sequoia Training Data, 2025

Cut burn, focus, bet

Near-death forced the pivot. A small war-room team built Agent while the rest of the company doubted it.

May 2024Laid off about 20% of staff to cut burn and focus.

Late 2024Overcame his distrust of sales and hired an enterprise sales leader.

“The worst part is when your team loses faith in your vision.”My First Million, 2026

Go direct, whatever it costs

Communicates personally and publicly, and holds positions even when they carry a price.

2023–26Outspoken on Gaza on X; says some investors distanced themselves.

Feb 2026Filed a First Amendment suit after a congressman blocked him on X.

“Contrarianism in Silicon Valley is largely fake.”SF Standard, 2026

Evolution

How the thinking changed

YearsEraWhat he believedWhat changed it

1993–2011

Amman

Systems can be bent by whoever understands them.

Changed his university grades via SQL injection, was caught, then helped secure the system.

2011–2016

Codecademy, Facebook

“Learn to code” is the answer.

Founding engineer at Codecademy; at Facebook helped start React Native and worked on Babel and Jest.

2016–2022

Replit, the classroom IDE

Move people from consuming software to creating it.

Co-founded with Haya Odeh. Loved by students, small in revenue. Early AI features in 2022.

2023–mid 2024

Near-death

Valued at $1.16B but stuck near $2.5M ARR.

Layoffs, and a team losing faith. He bet the company on agents.

Sep 2024–2025

Agent breakout

Anyone can vibe-code software. Stop telling people to learn to code.

Models got good at long-horizon tasks. ARR went from about $2.5M to $150M in a year.

2026

Knowledge-work platform

Fewer apps, more agents. Every knowledge worker builds.

Agent 4 (March 2026) runs parallel agents for apps, sites and decks. $9B valuation, targeting $1B ARR.

Scale

The numbers behind the posture

$250M

ARR by May 2026, from about $2.5M before Agent

$9B

Valuation on a $400M round, March 2026

~70

Employees at $150M ARR, October 2025

50M+

Users, with 85% of the Fortune 500 represented (2026)

$1M

ARR added on Agent’s launch day, September 2024

20%

Of staff cut in May 2024, months before the breakout

Tensions

Where the model strains

“I no longer think you should learn to code”

Also says minimal coding knowledge is gaining value

AI augments people rather than replacing them

A 70-person company; predicts smaller firms and layoffs

Criticizes “one prompt, build anything” hype

Replit’s own launches promise a startup from a prompt

Pushes long-running autonomous agents

The 2025 production-database deletion

Feels pushed out of Silicon Valley’s circles

Still backed by a16z, YC and Craft

For operators

What to take from it

When the core stalls, fund one frontier bet with a small team and protect it.

Hire for judgment and agency. Measure revenue per employee, not headcount.

Say what is broken in your product before your customers have to.

Audit the skills and beliefs you are proudest of. Drop the ones the facts have passed by.

Sources

Joe Rogan Experience #2344 (Jul 2025) · Sequoia Training Data (2025) · Lenny’s Podcast (2024) · a16z (2025–26) · My First Million (2026) · SaaStr AI (2026) · Platformer (Aug 2026) · SF Standard (2025–26) · Vanta (2025) · Replit blog (Mar 2026) · TechCrunch and Bloomberg (2025) · SiliconANGLE (May 2024) · amasad.me “Decade” essay (2021) · ADC case filing (2026). Quotes shortened.

All profilesAndy Jassy

Founder of AWS, CEO of Amazon · 1997 to 2026

Andy
Jassy

The Primitives Builder

He breaks every problem into basic building blocks, puts small teams of builders on them, and treats speed as a leadership choice. Now he is running a 1.5-million-person company as if it were a startup.

Drawn from his shareholder letters (2021 to 2025), his 2024 and 2025 memos, re:Invent keynotes, HBR IdeaCast (2025), CNBC, GeekWire, TechCrunch and earnings calls through July 2026.

The central method

Think in primitives

In 2003 he noticed every Amazon team was rebuilding the same storage, compute and databases. The fix was to sell the basic parts. Twenty years later he drew generative AI the same way: three layers, each a primitive others build on.

Swipe sideways to see the full diagram

2006 · The first primitives

S3 storage13 builders

EC2 compute11 builders

Tiny teams, per his 2024 letter. AWS is now a $169B run-rate business.

2024 · The same idea applied to AI

ApplicationsAlexa rebuilt on generative AI

Models and toolsBedrock

Chips and computeTrainium, 30 to 40% better price-performance

From the 2023 shareholder letter (April 2024). The bottom layer is where he is spending: $220B of 2026 capex.

Operating system

Eight principles

Break it into primitives

Find the foundational pieces everyone keeps rebuilding, and make those the product.

2003As Bezos’s technical advisor, traced slow software delivery to teams duplicating infrastructure. AWS followed.

2024Framed generative AI as chips, models and applications.

“Primitives are the raw parts or the most foundational-level building blocks.”2023 shareholder letter

The world’s largest startup

Size is not an excuse for slowness. He manages Amazon against the standard of a young company.

Sep 2024Ordered five days a week in the office to restore pace and collaboration.

Oct 2025Explained 14,000 corporate cuts as a culture move, not a cost move.

“Operating like the world’s largest startup.”Q3 2025 earnings call

Remove the layers

Bureaucracy compounds quietly. He goes after it with ratios, mailboxes and cuts.

2024–25Targeted 15% more individual contributors per manager; a Bureaucracy Mailbox drew ~1,000 emails and 375+ fixes.

Jan 202616,000 more roles cut to reduce layers and increase ownership.

“Pre-meetings for the pre-meetings for the decision meetings.”September 2024 memo

Speed is a choice

Leaders decide how fast an organization moves. Reversible decisions belong with the people doing the work.

2020re:Invent: reinvent while you are healthy, not when you are forced to.

2025Pushed two-way-door decisions down to builders.

“Speed is a leadership decision.”HBR IdeaCast, 2025

Small teams of builders

The unit of progress is a few builders who own the problem. Launch is the start, not the finish.

2006S3 built by 13 people, EC2 by 11.

2026Bedrock’s inference engine rebuilt in 76 days by six engineers using AI coding agents.

Builders, in his 2021 letter, are the people who dissect what does not work well and fix it.2021 shareholder letter

Experience is a moat

Operating at scale for years builds knowledge rivals cannot shortcut. Transformation takes multiple years.

2017Argued AWS’s head start would outlast newer cloud rivals.

2022Kept long-term bets through the downturn while cutting elsewhere.

“There is no compression algorithm for experience.”CNBC, 2017

Rebuild what works

When the underlying technology shifts, he rebuilds successful products from scratch rather than patching them.

2025Alexa rebuilt around generative AI.

2026Doubled his AWS outlook to at least $600B in revenue by 2036.

His 2025 letter argues reinvention is continuous, and that growth never runs in a straight line.2025 shareholder letter

Own the silicon

The economics of cloud AI are set at the chip. Custom silicon is how AWS keeps its margins and its customers.

2025Trainium2 pitched at 30 to 40% better price-performance than GPUs.

2026Anthropic deal for up to 5 GW of Trainium capacity.

“Demand for Trainium is booming.”2025 shareholder letter

Evolution

How the thinking changed

YearsEraWhat he believedWhat changed it

1997–2003

Apprentice

Learn by being in every room. He joined as a marketing manager and became Bezos’s shadow.

Saw the same infrastructure being rebuilt across teams, and guessed other companies had the same problem.

2003–2006

AWS founder

Amazon can sell its infrastructure as an operating system for the internet.

A 2003 six-page vision memo, then S3 and EC2 in 2006.

2006–2021

Evangelist and scaler

The cloud is still early. In 2020 it was roughly 4% of IT spend.

Became AWS CEO in 2016. Reinvention became his re:Invent theme as rivals arrived.

2021–2023

CEO in retrenchment

Keep the long-term bets, cut everything else.

Post-pandemic overhiring and a downturn. 27,000 corporate roles cut; AWS helped customers lower bills.

2024–2025

Largest startup

Layers and bureaucracy have crept in. AI makes lean teams possible.

Office five days a week, manager ratios, the mailbox, and a June 2025 memo saying AI will mean fewer people in some jobs.

2025–2026

All in on AI infrastructure

AI will reinvent every customer experience. Demand is ahead of capacity.

Capex raised to $220B for 2026, 30,000 more corporate cuts, a deeper Anthropic partnership.

Scale

The numbers behind the posture

$220B

Amazon’s planned 2026 capex, raised from $200B

$42.2B

AWS revenue in Q2 2026, up 37%, its fastest in 18 quarters

$600B

His AWS revenue outlook for 2036, doubled from $300B (March 2026)

57,000

Corporate roles cut across 2022–23 and Oct 2025–Jan 2026

13 + 11

Builders who shipped S3 and EC2 in 2006

$11B

Amazon free cash flow after the AI build, down from $38B

Tensions

Where the model strains

Oct 2025: the cuts are “not even really AI-driven”

June 2025 memo: AI will shrink the corporate workforce

Lean and scrappy

The largest capex budget in business, free cash flow down to $11B

Wants people to think like owners

Mandates office days and manager ratios from the top

No compression algorithm for experience

76-day rebuilds and “speed is a choice”

A neutral platform offering model choice

A large equity stake in Anthropic

For operators

What to take from it

Find what every team keeps rebuilding. That duplication may be your next product.

Open a bureaucracy mailbox and publish what you fix.

Push reversible decisions to the people doing the work, and say so explicitly.

Know which layer of your stack sets your economics, and own it.

Sources

Amazon shareholder letters (2021–2025) · Return-to-office memo (Sep 2024) · Generative AI memo (Jun 2025) · re:Invent keynote (2020) · HBR IdeaCast (May 2025) · CNBC (2017) · TechCrunch (2016) · GeekWire (2018, 2025, 2026) · Fortune (Jul 2026) · Futurum (Feb 2026) · SiliconANGLE (Mar 2026) · Amazon on Anthropic. Quotes shortened; paraphrases are marked.

All profilesSatya Nadella

CEO of Microsoft · 1992 to 2026

Satya
Nadella

The Refounder

Not the founder, but he rebuilt Microsoft as if he were. He changed the culture first, killed sacred cows, partners where he cannot win, and judges AI by whether it shows up in the real economy.

Drawn from his book Hit Refresh (2017), his memos (2014, 2023, 2025), Dwarkesh Patel (2025), BG2, Stratechery (2026), Pivot, CNN GPS, Davos and Microsoft earnings through July 2026.

The central pattern

Reframe, then rebuild

Each phase of his tenure starts with a new sentence about what Microsoft is. The frame comes first; reorganizations, acquisitions and writedowns follow from it.

Swipe sideways to see the full diagram

2014Mobile-first, cloud-first

Windows stops being the center. Linux embraced, Office on iPhone, Nokia written off.

2017Intelligent cloud, intelligent edge

Developers and open source as the platform. LinkedIn, then GitHub.

2019–23AI is the next platform

Bet on OpenAI and become the infrastructure underneath it.

2025–26Tokens per dollar per watt

A capital-intensive business. Models are commodities; power is the constraint.

Market value across the arc $3T Jan 2024 $4T Jul 2025

Operating system

Eight principles

Learn-it-all beats know-it-all

Borrowed from Carol Dweck’s growth mindset. Curiosity over looking clever, applied to the whole company.

2014–17Moved Microsoft from a culture where people competed to seem smartest to one built on learning.

2025Told staff to keep rapidly learning new skills as AI reshapes roles.

“For too long we’ve been the know-it-all company.”On Hit Refresh, Next Big Idea Club

Empathy as a source of product

Shaped by raising his son Zain, who had cerebral palsy. Good products meet needs customers cannot articulate.

2017Described fatherhood as something he had to rise to, and credits it with teaching him empathy.

2017Tied Microsoft’s best products to unmet, unarticulated needs.

“Empathy make[s] you a better innovator.”Fortune, 2017

The CEO curates culture

Strategy follows culture. He treats the job as shaping how people think, one person at a time.

Feb 2014First email as CEO asked Microsoft to rediscover its soul.

2017Hit Refresh made the culture change the official story.

“Curator of an organization’s culture.”Hit Refresh, via Fast Company

Kill sacred cows

Loyalty to past strategies is a liability. He dismantles them quickly once the frame changes.

Oct 2014“Microsoft loves Linux,” with 20% of Azure already running it.

Jul 2015$7.6B Nokia writedown and 7,800 job cuts, effectively ending Windows Phone.

“Our industry does not respect tradition.”First email as CEO, 2014

Partner where you can’t win

Be a platform others can trust rather than trying to own every category.

2018Bought GitHub for $7.5B and promised to keep it open and independent.

2025Built a fungible data center fleet able to run many families of models.

“Judge us by the actions we have taken.”To GitHub developers · The Register, 2018

Control your own destiny

After the OpenAI board crisis, he hedged: rights to the models, an in-house lab, and multiple providers.

Nov 2023Moved to protect the partnership during Sam Altman’s firing and return.

2024–26Hired Mustafa Suleyman to build Microsoft’s own models; warns firms against relying on one AI.

“Never going to get back into a situation where we get surprised.”Pivot, November 2023 (shortened)

Judge AI by outcomes

Declaring AGI is benchmark theater. The test is economic growth and whether society lets AI use scarce energy.

Feb 2025Called self-declared AGI milestones nonsensical benchmark hacking.

Jan 2026At Davos, warned AI will lose social permission unless its benefits spread.

“The real benchmark is: the world growing at 10%.”Dwarkesh Patel, 2025

Think in unit economics

Microsoft is now capital-intensive. The value sits in the scaffolding around models, and the limit is power.

Nov 2025Said he had chips in inventory he could not plug in for lack of power.

Oct 2025Handed the commercial business to Judson Althoff to focus on data centers, systems and AI.

“Tokens-per-dollar-per-watt.”Stratechery, 2026

Evolution

How the thinking changed

YearsEraWhat he believedWhat changed it

1992–2013

Engineer and builder

Enterprise infrastructure matters more than the Windows franchise.

Ran Bing, then Server and Tools, where Azure grew. A front-row view of the cloud inside a Windows company.

2014–2016

Refounder

Microsoft missed mobile; the cloud is the next chance.

Linux, Office on rival platforms, the Nokia writedown and LinkedIn for $26.2B.

2017–2018

Culture made official

Developers and open source are the platform.

Hit Refresh codified the learn-it-all culture; GitHub confirmed the developer bet.

2019–2022

The OpenAI bet

Scaling will create a new platform; be the infrastructure underneath it.

$1B in 2019, growing to about $13.8B invested, with Azure as OpenAI’s compute provider.

2023–mid 2025

AI-first, then hedged

AI is the next major wave of computing.

Activision closed ($69B). The November 2023 crisis exposed the dependency; he added Suleyman, in-house models and multi-model choice.

mid 2025–2026

Capital-intensive Microsoft

From software factory to intelligence engine. AI must earn its social permission.

The 2025 “enigma of success” layoffs, a restructured OpenAI deal (27% stake), and a CEO role narrowed to technology.

Scale

The numbers behind the posture

$4T

Market value in July 2025, the second company ever to reach it

$100B+

Azure revenue in fiscal 2026, its first year above the mark

27%

Microsoft’s OpenAI stake after the October 2025 restructuring, worth about $135B

$7.6B

Nokia writedown, July 2015: the sacred cow he killed first

$69B

Activision Blizzard, closed October 2023

15,000+

Jobs cut in 2025 while spending about $80B on AI capex

Tensions

Where the model strains

Empathy as the core of leadership

About 15,000 cut in a record-profit year

OpenAI as a true partner

Trial emails show him insisting on control of destiny

Models are commoditizing

Paid heavily to hire a team to build frontier models

AI must show up in GDP

Copilot adoption is still contested

Warns the industry is too self-obsessed

Runs one of the largest infrastructure buildouts in history

For operators

What to take from it

Write the one sentence that says what your company is now. Reorganize around it.

Name the strategy you are most attached to, and test whether it still earns its place.

Partner deeply, but never let a single partner hold your destiny.

Judge new technology by customer outcomes, not benchmarks.

Sources

Satya Nadella, Hit Refresh (2017) · First email as CEO (Feb 2014) · Microsoft memos (Jan 2023, Jul 2025, Oct 2025) · Fortune (2017, 2024, 2025) · The Register (2014, 2018) · TechCrunch (2015, 2026) · Pivot via GeekWire (Nov 2023) · Dwarkesh Patel (Feb and Nov 2025) · BG2 (2024, 2025) · Stratechery (Jun 2026) · CNN GPS (Jul 2026) · Davos (Jan 2026) · Microsoft FY26 Q4 results · Windows Central (2026). Quotes shortened.

All profilesNelson Mandela

Anti-apartheid leader, President of South Africa · 1944 to 2013

Nelson
Mandela

The Patient Negotiator

He held a few principles fixed and treated everything else as tactics. He studied his opponent until he could speak his language, moved ahead of his own side when the moment came, then brought them with him.

Drawn from his own words: Long Walk to Freedom (1994), Conversations with Myself (2010), the Rivonia statement (1964), prison letters and major speeches; plus Richard Stengel’s Time essay on his lessons of leadership (2008).

The central pattern

Fixed goal, flexible means

The goal never moved: a non-racial democracy. The methods changed whenever the old ones stopped working, including the ones his admirers would rather forget.

Swipe sideways to see the full diagram

Fixed: one person, one vote, a non-racial South Africa1944 to 1999

1952

Mass defiance

Nonviolent campaigns against pass laws and segregation.

1961

Armed struggle

After Sharpeville and the ANC ban, co-founded the armed wing, beginning with sabotage.

1964–84

Prison as school

Learned Afrikaans and Afrikaner history; studied his jailers.

1985

Secret talks

Wrote to the government proposing talks, without his movement’s approval.

1994

Reconciliation

A unity government, the Truth Commission, the Springbok jersey.

Flexible: the method changed five times in fifty years. The goal did not.

Operating system

Eight principles

Principles fixed, tactics not

Nonviolence was a tactic to him, not a creed. He changed methods when the state closed every legal route, and would not trade principle for his own freedom.

1961After Sharpeville and the ANC ban, co-founded Umkhonto we Sizwe, starting with sabotage.

1985Refused release offered on condition that he renounce violence.

“Unrealistic and wrong for African leaders to continue preaching peace and non-violence.”Rivonia trial statement, 1964 (shortened)

Move ahead of the flock

When he judged the moment right, he acted before his movement authorized it, then won colleagues over one at a time.

1985From prison, wrote to Justice Minister Kobie Coetsee proposing talks.

1985–89Brought fellow prisoners and the exiled leadership along individually.

“There are times when a leader must move out ahead of the flock.”Long Walk to Freedom, 1994

Lead from behind

The model he learned at the regent’s court: hear everyone, speak last, and let the group own the decision.

BoyhoodWatched the regent listen to every voice before summing up.

1990In his release speech, placed decisions with the movement’s elected structures.

The best leaders guide people who are “not realizing they are being led from behind.”Long Walk to Freedom, 1994

Master yourself first

Composure was a discipline and a tool. He showed calm in public whatever he felt in private.

1975Wrote from Robben Island that a cell is a place to come to know yourself.

1994Stayed calm when a campaign plane’s engine failed, then admitted he had been terrified.

“You must put up a front.”To Richard Stengel · Time, 2008

Study the opponent

He learned his jailers’ language, history and sport, because negotiation starts with understanding what the other side fears.

1960s–70sLearned Afrikaans on Robben Island and helped warders with their legal problems.

1995Wore the Springbok jersey to the Rugby World Cup final at Ellis Park.

“Sport has the power to change the world.”Laureus World Sports Awards, 2000

Keep rivals close

He brought opponents inside the tent, sharing power and credit so the transition would not break.

1993After Chris Hani’s assassination, gave the national address the government could not.

1994–96Governed with F.W. de Klerk as a deputy president; set up the Truth and Reconciliation Commission.

“A white woman, of Afrikaner origin, risked her life so that we may know.”Address after Hani’s assassination, 1993

Symbols are strategy

He understood a gesture can move a country faster than a policy.

1994Staged his inauguration as a moment of national healing.

1995The jersey and handshake signalled to white South Africans that he held no bitterness.

“The time for the healing of the wounds has come.”Inauguration address, 1994

Know when to leave

He served one term and stepped down, a precedent many liberation leaders never set.

1996Confirmed he would not seek a second term.

1999Handed over power and turned to elder work, including AIDS campaigning.

Cyril Ramaphosa on his role: “His job was to set the course, not to steer the ship.”Time, 2008

Evolution

How the thinking changed

YearsEraWhat they believedWhat changed it

1944–1951

Youth League militant

African nationalism, wary of alliances with communists and Indian activists.

Through Walter Sisulu and joint campaigns, moved toward a non-racial alliance.

1952–1964

Defiance to armed struggle

Mass nonviolent defiance could force change.

Sharpeville and the ANC ban closed legal routes. He co-founded the armed wing and was sentenced to life at Rivonia.

1964–1984

Prisoner as student

Endure, organize, learn.

Daily contact with warders; he stopped seeing Afrikaners only as the enemy.

1985–1993

Negotiator

Only negotiation can break the stalemate.

Refused conditional release, opened secret talks, walked free in 1990, held the process together after Hani’s murder.

1994–1999

President as reconciler

A durable constitution matters more than a personality.

Unity government, the Truth Commission, the rugby jersey, then one term.

1999–2013

Elder

Moral authority is best spent on what was neglected.

Campaigned openly on HIV/AIDS and announced his son’s death from it in 2005.

Scale

The numbers behind the posture

Years in prison, 1962 to 1990

Years into his sentence when he opened talks with the government, 1985

63%

ANC share of the vote in the first democratic election, April 1994

His age at inauguration, May 1994

Term served as president, 1994 to 1999

849

Amnesties granted by the Truth Commission, out of 7,111 applications

Tensions

Where the model strains

An icon of peace

Founded an armed wing and would not renounce it until negotiations were secured

A humble servant of the movement

Opened talks alone and told colleagues afterwards

Political reconciliation

Critics argue apartheid-era economic structures stayed largely intact

Moral authority on every issue

Largely quiet on AIDS early in his presidency as infection rates rose

Rejected sainthood himself

Remembered as a saint, which flattens the costs borne by his family

For operators

What to take from it

Write down the two or three things you will never trade. Treat everything else as tactics.

When you move ahead of your team, go back for them one person at a time.

Learn your counterpart’s language before you negotiate with them.

Decide your exit before you need one.

Sources

Nelson Mandela, Long Walk to Freedom (1994) and Conversations with Myself (2010) · Rivonia trial statement (1964) · Reply to P.W. Botha (1985) · Release speech (1990) · Address after Chris Hani’s assassination (1993) · Inauguration address (1994) · Richard Stengel, Time (2008) · Harvard Program on Negotiation · PBS Frontline · Nelson Mandela Foundation · Laureus (2000). Quotes shortened.

All profilesNapoleon Bonaparte

General, First Consul, Emperor of the French · 1769 to 1821

Napoleon
Bonaparte

The Tempo Operator

He found the decisive point and put everything there, faster than anyone thought possible. He built institutions that outlived him, and never set himself a stopping rule.

Drawn from his own correspondence (over 40,500 letters, Fondation Napoléon), Las Cases’ Mémorial de Sainte-Hélène, and the biographies of Andrew Roberts (2014) and Adam Zamoyski (2018). Many famous Napoleon quotes are apocryphal and are left out.

The central pattern

Speed, then overreach

The tempo that won Italy and Austerlitz carried him into Russia. Charles Minard’s 1869 chart of the 1812 campaign is the classic picture of a winning method with no stopping rule.

Swipe sideways to see the full diagram

Main body only, band width proportional to men, after Charles Minard (1869). His own line in Warsaw that December: “From the sublime to the ridiculous is but one step.”

Operating system

Eight principles

Find the decisive point

He looked for the single position that decides everything, then massed force there.

1793At Toulon, saw that the forts above the harbour controlled it; his guns forced the British fleet out. Brigadier general at 24.

1805At Austerlitz, left the Pratzen Heights open to draw the Allies in, then struck their weakened centre.

“One bad general is better than two good ones.”Letter to the Directory, 1796, against splitting command

Time is the scarcest resource

Tempo was his edge: strike before the enemy can concentrate.

1796Won three battles in about two weeks and knocked Piedmont out before Austria could combine.

1805Marched seven corps from the Channel to the Danube in about five weeks and trapped an Austrian army at Ulm.

“Activité, activité, vitesse!”Letter to Masséna, 1809

Central standard, local execution

Self-contained units that can act alone, joined by one clear chain of command.

1800Created prefects, one per department, reporting to Paris.

1804–05The corps system: each corps a small army able to hold a larger force until the others arrived.

A corps, in his design, could fight alone for a day while the rest of the army marched to it.Roberts, Napoleon the Great (paraphrase)

Beat them one at a time

Get between two opponents and defeat each before they can combine.

1796In Italy, split the Austrian and Piedmontese armies and beat them separately.

1815Drove between Wellington and Blücher and won at Ligny, then lost at Waterloo when the Prussians arrived.

Later strategists named it the central position, and built manuals around it.After Jomini (paraphrase)

Morale is a multiplier

He spoke to soldiers directly, ran his own newspapers and invented honours anyone could earn.

1797Founded army newspapers in Italy.

1802Created the Legion of Honour for soldiers and civilians alike.

“It is with baubles that men are led.”Council of State, 1802, recorded by Thibaudeau

Merit and legible rules

Careers open to talent, and one code of law anyone could read.

1804The Civil Code: 2,281 articles on equality before the law, property and civil status.

1804Named marshals including an innkeeper’s son (Murat) and a cooper’s son (Ney).

On Saint Helena he said his Code, not his forty victories, would last.Recorded by Montholon (paraphrase)

Own the narrative

Communication was part of operations. He shaped his legend as it happened.

1796 onArmy bulletins so self-serving that “to lie like a bulletin” became a French saying.

1798Brought scholars to Egypt, making the expedition about learning as well as conquest.

“What a novel my life has been!”Las Cases, Mémorial de Sainte-Hélène

Know the details yourself

A prodigious memory and work rate. He held positions, officers and supplies in his head.

1784–1821Wrote or dictated more than 40,500 letters, on everything from forage to the Opéra.

CampaignsKnew unit locations, officers’ names and supply figures without notes, per Roberts.

His letters cover roads, troop numbers and theatre casting in the same week.Fondation Napoléon (paraphrase)

Evolution

How the thinking changed

YearsEraWhat they believedWhat changed it

1769–1793

Corsican outsider

Corsican nationalism, with Pasquale Paoli as his hero.

The family broke with Paoli and fled to France. He committed to the French state.

1793–1796

Technical specialist

Competence with guns and maps wins.

Toulon made his name; in 1795 he crushed a royalist rising in Paris with cannon.

1796–1799

Autonomous general

He decides, not the government.

Negotiated peace on his own authority, invaded Egypt, then left his army there for Paris.

1799–1804

Institution builder

Order, merit and religious peace make a regime legitimate.

The 1799 coup, prefects, the Banque de France, the Concordat, the Legion of Honour, the Civil Code.

1804–1812

Emperor-conqueror

Victory is his legitimacy.

Austerlitz, Jena, Tilsit, then the Continental System, Spain in 1808 and Russia in 1812.

1814–1821

Exile and myth-maker

Posterity is the audience.

Abdication, Elba, the Hundred Days, Waterloo, then Saint Helena, where he dictated his legend.

Scale

The numbers behind the posture

Battles fought, of which 7 lost (Roberts)

His age when made brigadier general after Toulon, 1793

40,500+

Letters, 1784 to 1821, now fully published

2,281

Articles in the Civil Code, 1804

422,000

Main body that entered Russia in June 1812; about 10,000 came back (Minard)

500,000+

Troops engaged at Leipzig, 1813, Europe’s largest battle before 1914

Tensions

Where the model strains

Careers open to talent

Made his brothers kings of Naples, Holland and Westphalia

Delegated through the corps system

Marshals trained to execute, not judge, faltered without him

Speed as his edge

No stopping rule: the Continental System led to Spain, then Russia

Legal equality in the Code

Restored slavery in the colonies in 1802, censored the press, subordinated women in law

Master of propaganda

By 1812 his information system told him what he wanted to hear

For operators

What to take from it

Find the one position that decides the outcome, and concentrate there.

Build units that can operate alone for a day, joined by one chain of command.

Set a stopping rule before the winning streak starts.

Codify what you build. Institutions outlast campaigns.

Sources

Fondation Napoléon, General Correspondence (2004–2018) · Las Cases, Mémorial de Sainte-Hélène (1823) · Andrew Roberts, Napoleon the Great (2014) · Adam Zamoyski, Napoleon (2018) and 1812 (2004) · Charles Minard (1869) · Britannica · Tim Ferriss #773 and Dwarkesh Patel with Andrew Roberts · Blue Ocean Strategy teaching notes. Apocryphal quotes excluded; paraphrases marked.

All profilesBrian Chesky

Co-founder and CEO of Airbnb · 2007 to 2026

Brian
Chesky

The Founder in the Details

An industrial designer who ended up running a public company, and refused to stop designing. He meets users by hand, imagines past the reasonable, and answers scale by staying close to the work rather than delegating away from it.

Drawn from Masters of Scale (2017), Stanford GSB (2023), Config (2023), Lenny’s Podcast (2023), his May 2020 letter to employees, his 2024 Y Combinator talk and Paul Graham’s “Founder Mode,” and interviews with Fortune, Skift and TechCrunch (2024 to 2026).

The central method

The 11-star experience

His signature design exercise: define a 5-star stay, then imagine 6, 7, all the way to an absurd 11. Somewhere between the fantasy and the ordinary is a version you can actually build.

Swipe sideways to see the full diagram

Broken

Expected

Remarkable and buildable

Fantasy, useful for stretching

Paired with Paul Graham’s advice at Y Combinator in 2009: do things that don’t scale. The founders went to New York hosts’ homes and photographed the listings themselves.

Operating system

Eight principles

Do things that don’t scale

Meet users one by one, fix problems by hand, then design the system.

2009Flew to New York with Joe Gebbia to meet hosts and photograph their listings himself.

2020After COVID, cut marketing from nearly $1B to zero and went back to the core host product.

“Get your customers one by one.”Paul Graham’s advice, as Chesky retold it · Masters of Scale, 2017

Design past reasonable

Imagine far beyond what is sensible, then walk back to what is buildable and remarkable.

2017Described the 11-star exercise on Masters of Scale.

2025Relaunched Services and Experiences to design the whole trip, not only the stay.

“What would an 11-star experience be?”Masters of Scale, 2017

The founder in the details

The CEO reviews the work directly instead of managing through layers.

2023Moved Airbnb to a functional organization with one roadmap that he reviews.

2024His talk at Y Combinator prompted Paul Graham’s essay “Founder Mode.”

“Great leadership is presence, not absence.”Fortune, 2026

Close, not layered

He keeps a wide circle of people reporting straight to him so nothing important is filtered.

2025Said he treats about 50 people as direct reports.

Sep 2024Pushed back on the label: he never called it founder mode, and it isn’t micromanagement.

“I never called it founder mode.”Skift, 2024

One product, one story

Product and marketing are one job; design sits beside engineering as an equal.

2023Merged classic product management into product marketing.

May 2025Launched the redesigned app himself in a single keynote.

“We got rid of the classic product management function.”Config, 2023

Cut to the core in a crisis

When the business collapsed, he narrowed it to what only Airbnb could do.

Spring 2020Watched bookings fall about 80% in eight weeks.

May 2020Paused transportation and film projects and scaled back hotels and luxury to focus on hosts.

“The hardest thing to manage in a crisis is your own psychology.”Stanford GSB, 2023

Care on the way out

How you let people go is part of the culture you keep.

May 2020Laid off 1,900 people with 14 weeks’ severance plus a week per year, and waived the one-year equity cliff.

May 2020Explained the reasoning in a long letter to every employee, published in full.

His letter told those leaving that the cuts were no reflection on their work.Airbnb newsroom, May 2020 (paraphrase)

AI on proprietary ground

Adopt AI fast, but build it on what competitors can’t copy: identity, reviews and supply.

Feb 2026AI handles about a third of North American customer support.

Aug 2026Around 60% of new code is written by AI.

“A chatbot doesn’t have our 200 million verified identities.”TechCrunch, 2026

Evolution

How the thinking changed

YearsEraWhat they believedWhat changed it

2004–2007

Industrial designer

Design is how you make things people love.

Graduated from the Rhode Island School of Design, worked as a designer in Los Angeles, then moved to San Francisco in 2007.

2007–2009

Scrappy founder

People will pay to sleep on an air mattress in a stranger’s home.

Most investors passed. Selling cereal and getting into Y Combinator kept the company alive.

2010–2014

Learning to be CEO

Belonging is the product, and trust has to be designed.

A host’s home was trashed in 2011; trust and safety became a core system; he led the 2014 rebrand.

2015–2019

Expansion

Airbnb can become a full travel company.

Experiences, China, luxury and hotels created divisional sprawl and regulatory fights.

2020

Near-death

Focus over breadth.

Bookings fell 80% in eight weeks; a quarter of staff laid off; IPO in December anyway.

2021–2026

Founder mode

The CEO is the chief product officer, and AI is the next platform.

One roadmap and twice-yearly releases, then Services and Experiences, and AI across support and code.

Scale

The numbers behind the posture

~50

People he treats as direct reports, 2025

−80%

Bookings in eight weeks at the start of the pandemic, 2020

1,900

People laid off, about a quarter of staff, May 2020

$146

First-day opening price against a $68 IPO price, December 2020

$11.1B

Airbnb revenue in 2024

60%

Share of new code written by AI, August 2026

Tensions

Where the model strains

Presence and founder mode

Critics see one person as the bottleneck; he rejects the micromanager label

Built on doing things that don’t scale

A 2026 plan to automate support and code

2020: cut to the core

2025: expanded into services and experiences again

“Belong Anywhere”

Cities restricting short-term rentals amid housing shortages

For operators

What to take from it

Do the unscalable version by hand first. It tells you what to automate.

Run the 11-star exercise on your core flow, then build the 7.

Stay close to the work. Layers are where quality leaks.

In a crisis, cut to the core, and be generous with the people you let go.

Sources

Masters of Scale (2017) · Airbnb newsroom, letter to employees (May 2020) · Stanford GSB (2023) · Config (2023) · Lenny’s Podcast (2023) · Skift (2024) · Paul Graham, “Founder Mode” (2024) · Fortune (2025, 2026) · TechCrunch (2020, 2026) · CNBC (2020) · Stratechery (2025). Quotes shortened; paraphrases marked.

All profilesBob Iger

CEO of The Walt Disney Company · 1974 to 2026

Bob
Iger

The Brand Collector

He bought creative engines instead of building them, repaired relationships to get the deals done, and chose to disrupt Disney’s own business before someone else did.

Drawn from his memoir The Ride of a Lifetime (2019), Tim Ferriss (2020), NPR (2019), DealBook (2023), interviews with CNBC, Fortune and Variety, his 2022 return memo and Disney’s 2026 succession announcement.

The central strategy

Buy the creative engines

Instead of trying to out-create Pixar, Marvel and Lucasfilm, he bought them and protected their cultures. Then he bought scale for streaming, and the debt that came with it.

Swipe sideways to see the full diagram

Pixar2006

$7.4B

Marvel2009

$4B

Lucasfilm2012

$4.05B

21st Century Fox2019

$71.3B

Purchase prices, bars to scale. Disney+ launched in November 2019, eight months after the Fox deal closed.

Operating system

Eight principles

Disrupt yourself first

Move before disruption forces you to, even at a short-term cost.

2005Named three priorities as CEO: quality content, technology, going global.

2019Launched Disney+, knowing it would cannibalize licensing income.

“If you don’t innovate, you die.”The Ride of a Lifetime, 2019

Courage over safety

Long shots are shorter than they look when the upside is large.

2006Pitched Steve Jobs on Disney buying Pixar.

2012Bought Lucasfilm from George Lucas.

“Don’t be in the business of playing it safe.”The Ride of a Lifetime, 2019

Stay out of small markets

Advice from his Cap Cities boss Dan Burke that he kept in his desk.

1985–95Learned lean, decentralized management under Tom Murphy and Dan Burke.

2023Said ABC and other linear networks “may not be core.”

“Avoid getting into the business of manufacturing trombone oil.”Dan Burke, quoted in The Ride of a Lifetime

Decide, and solve bad news

Chronic indecision costs more than a wrong call.

2005 onApproached every negotiation intending to close.

Feb 2023Announced $5.5B in cuts and 7,000 layoffs about 11 weeks after returning.

“I tend to approach bad news as a problem that can be worked through.”The Ride of a Lifetime, 2019

High standards, humanely

Roone Arledge taught him not to accept mediocrity. He tried to keep the standard without the cruelty.

1970s–80sWorked at ABC Sports under Arledge.

2023Admitted Marvel had lost quality through volume.

“Quantity, in our case, diluted quality.”DealBook, 2023

Relationships close deals

Candor and respect with founders made the acquisitions possible.

2005–06Rebuilt Disney’s relationship with Steve Jobs after years of feuding.

2009–12Won Marvel and Lucasfilm on personal trust with their owners.

“Empathy is a prerequisite to the sound management of creativity.”The Ride of a Lifetime, 2019

Own your mistakes

Authenticity includes naming what you got wrong, in public.

Feb 2023Said success had made him dismissive of other people’s ideas.

Sep 2026Said the 2020 succession process may not have been thorough enough.

“I got overconfident in my own instincts.”Fortune, 2023

Put creators back in charge

Decisions about creative work belong with creative teams, not a central planning office.

2005Dismantled the strategic planning group he inherited.

2023Returned decision-making to Disney’s creative teams.

“More decision-making back in the hands of our creative teams.”Fortune, 2023

Evolution

How the thinking changed

YearsEraWhat they believedWhat changed it

1974–1985

ABC Sports

Excellence comes from Arledge-style demands.

Learned showmanship, and that no is rarely final.

1985–1995

Cap Cities/ABC

Lean, decentralized, decent management.

Tom Murphy and Dan Burke promoted him beyond his experience to run ABC Entertainment.

1996–2005

Number two to Eisner

Patience and loyalty under a difficult boss.

Watched feuds with Jobs and Roy Disney and learned what not to do at the top.

2005–2019

Empire building

Buy creative engines; brands win in a disrupted world.

Pixar, Marvel, Lucasfilm, Fox, Disney+.

2020–2022

Step-down

Believed succession was handled.

His successor centralized control, the stock fell about 40% in 2022, and the board called him back.

2022–2026

Return and exit

Humility, creative autonomy, lower costs.

$7.5B in savings, won the 2024 proxy fight, named Josh D’Amaro CEO from March 2026; retires at the end of 2026.

Scale

The numbers behind the posture

$71.3B

21st Century Fox, closed March 2019

$7.4B

Pixar, announced January 2006

$4B

Marvel, 2009, followed by Lucasfilm for $4.05B in 2012

$7.5B

Cost savings target after his return, November 2023

7,000

Jobs cut in the 2023 restructuring

2026

Handed over to Josh D’Amaro in March; retires December 31

Tensions

Where the model strains

Preaches preparing successors

Extended his contract repeatedly, picked a successor, then returned

Empathy and thoughtfulness

7,000 layoffs, and calling 2023 strike demands “not realistic”

The status quo is the real risk

The Fox price left a heavy balance sheet behind the Peltz fight

Built the franchise machine

Later said Disney had made too many sequels

Decentralized creative power

By his own account, dismissive of other people’s ideas

For operators

What to take from it

Buy capability you can’t build in time, then protect its culture.

Kill your own cash cow before a competitor does.

Name your successor early, and test them in a crisis.

Say what you got wrong, specifically.

Sources

Robert Iger, The Ride of a Lifetime (2019) · Tim Ferriss #406 (2020) · NPR (2019) · ABC News book excerpt (2019) · Fortune (2022, 2023) · CNBC (2023, 2024, 2026) · DealBook via CNN (2023) · Variety (2026) · Disney press release (February 2026) · Inc. · Forbes (2023). Quotes shortened.

All profilesJack Dorsey

Co-founder of Twitter and Block · 2006 to 2026

Jack
Dorsey

The Protocol Minimalist

He runs a company the way an editor treats a text: cut until only what matters remains. He has come to believe the most important things online should be protocols no company owns.

Drawn from TED (2019), Joe Rogan (2019), Ben Greenfield (2019), his 2021 resignation letter, Block shareholder letters and memos (2024 to 2026), the essay “From Hierarchy to Intelligence” (2026), and his posts on X and Nostr.

The central shift

From company to protocol

His biggest regret about Twitter is that it became a company at all. Since 2021 he has pushed toward systems nobody owns: bitcoin, Nostr, a messenger with no servers.

Swipe sideways to see the full diagram

2006Twitter2009Square2021Block, and leaves Twitter2022Funds Nostr2025Builds bitchat

Operating system

Eight principles

Protocols over platforms

No state or company should own the public conversation.

Aug 2022Called Twitter becoming a company his biggest regret.

2022–25Gave about 14 BTC to Nostr; built bitchat, a messenger with no servers, over a weekend.

“My biggest regret is that it became a company.”Post on X, 2022, via TechCrunch (shortened)

Bitcoin as the internet’s money

He expects the internet to have a native currency, and thinks it will be bitcoin.

2019Predicted it on Joe Rogan.

2022Told investors Block was no longer just a payments company; later built mining chips and a wallet.

“If I weren’t at Square or Twitter, I’d be working on bitcoin.”Bitcoin 2021, Miami

The CEO as editor

The job is to filter: choose which of many good ideas get made.

2010sDescribed the CEO as chief editor in a Stanford talk.

2015Square’s IPO letter: better a small part of something big.

“I’d rather have a smaller part of something big.”Square S-1 letter, 2015 (shortened)

Founders should be replaceable

A company that depends on its founder has a single point of failure.

Nov 2021Stepped down as Twitter CEO.

May 2024Left the Bluesky board over its direction.

Founder-led is “severely limiting and a single source of failure.”Twitter resignation letter, 2021

Flatten the organization

Layers slow information, so he removes them.

Mar 2025Cut 931 roles, removed 80 managers and capped org depth.

Mar 2026Proposed three roles only: builders, directly responsible individuals, player-coaches.

“Hierarchy and middle management impede information flow.”“From Hierarchy to Intelligence,” 2026

Cut early, on your own terms

Better to shrink by choice than be forced to.

Mar 2025Said Block was behind in acting, and that wasn’t fair to staff.

Feb 2026Cut more than 4,000 people, about 40% of Block.

“I’d rather get there honestly and on our own terms.”Layoff statement, 2026

Discipline of the body

Fasting, cold, meditation and walking as tools for focus.

2018Ten-day silent meditation retreat in Myanmar.

2019One meal a day, weekend fasts, a five-mile walk to work.

“Nothing has given me more mental confidence” than the cold plunge.Ben Greenfield podcast, 2019

Admit design mistakes

He says in public what he would build differently.

Feb 2019Conceded a live-tweeted interview was hard to follow.

Apr 2019Said he would not emphasize follower counts if starting again.

“I don’t think I would create ‘likes’ in the first place.”TED, 2019

Evolution

How the thinking changed

YearsEraWhat they believedWhat changed it

~1990–2005

Dispatch kid

The world as real-time status updates.

Dispatch routing software became the idea behind Twitter.

2006–2008

Twitter founder

Simplicity, constraint, craft.

Pushed out as CEO in 2008 when the board lost confidence in him as an operator.

2009–2015

Square, the editor

Design-led, with the CEO as editor.

Square rebuilt his credibility as a product builder.

2015–2021

Two public companies

He could run Twitter and Square at once.

Critics cited split focus. He pledged $1B to his Start Small fund in 2020.

2021–2024

Protocol purist

Social should be a protocol; money should be bitcoin.

Left Twitter, renamed Square as Block, backed then soured on Musk’s purchase, funded Nostr, left Bluesky.

2024–2026

AI-native Block

A company built as an intelligence, with speed over hierarchy.

Functional reorg, the goose AI agent, 931 cuts, then about 40% cut in 2026.

Scale

The numbers behind the posture

~40%

Of Block’s staff cut in February 2026, from over 10,000 to under 6,000

931

Roles cut in March 2025, including 80 managers

14 BTC

Given to fund Nostr, December 2022

$1B

Pledged to his Start Small fund, April 2020

$12.51B

Block’s raised 2026 gross profit guidance

5 miles

His walk to work, 2019

Tensions

Where the model strains

Nobody should own Twitter

Called Musk “the singular solution I trust” in 2022, then said “it all went south”

Founder-led is a single point of failure

Remains Block’s head and drove its restructuring himself

Bitcoin purist

Block added stablecoins because customers wanted them

March 2025: not replacing people with AI

February 2026: AI given as the reason to cut 40%

Funds many open projects

Has walked away from several: Bluesky, the Web5 unit, Tidal

For operators

What to take from it

Ask which part of your product should be a protocol nobody owns.

Act as editor: say no to good ideas so the great ones ship.

Measure org depth, and cap it.

If cuts are coming, make them early and explain why.

Sources

TED (2019) · Joe Rogan (2019) · Ben Greenfield (2019) · Recode (2019) · Square S-1 letter (2015) · Twitter resignation letter via Axios (2021) · Bitcoin 2021 · CNBC (2022) · TechCrunch (2022, 2025) · CoinDesk (2022, 2026) · Fortune (2024, 2025, 2026) · Block, “From Hierarchy to Intelligence” (2026) · Block Q4 2025 shareholder letter · Newsweek (2026). Quotes shortened.

All profilesEvan Spiegel

Co-founder and CEO of Snap · 2011 to 2026

Evan
Spiegel

The Camera Purist

He designs from the camera out, keeps friends separate from media, and protects long bets with founder control. Kindness is his stated culture; deep cost cuts are his repeated reality.

Drawn from his Stanford talk (2013), Columbia (2016), Snap’s S-1 (2017), his 2017 Axios op-ed, memos (2018 to 2026), Diary of a CEO (2025), Colin and Samir (2025), Lenny’s Podcast (2026) and Snap’s 14-year letter (2025).

The central design

The camera in the middle

Snapchat opens to the camera, not a feed. Friends sit on one side, media on the other, and he has spent a decade keeping them apart.

Swipe sideways to see the full diagram

ChatFriends. Private, disappearing by default.

←

Opens hereCameraEvery session starts by making something, not scrolling.

→

Stories and SpotlightMedia. Creators and publishers, kept separate from friends.

The same bet, moved onto the face: Specs augmented-reality glasses, preorders September 2026. $2,195

Operating system

Eight principles

Deleting is the default

Ephemerality lets people talk without performing for a permanent record.

2011Picaboo, built with Bobby Murphy around disappearing photos.

2014Framed Snapchat as talking through content rather than around it.

“Deleted everything by default and saved the things that are important to us.”Stanford talk, 2013

Independence over an early exit

Keep control so you can make decade-long bets.

2013Turned down Facebook’s offer of roughly $3B.

2017Took Snap public selling only non-voting shares.

“Trading that for some short-term gain isn’t very interesting.”Forbes, 2014

The camera is the product

Snap is a camera company; the app is one expression of it.

2016Renamed the company Snap Inc. and launched Spectacles.

2026Launched Specs augmented-reality glasses.

“Snap Inc. is a camera company.”Snap S-1, 2017

Friends are not media

Separate talking with friends from watching creators.

2017Argued for the split in Axios; the 2018 redesign followed.

2024Memo: “Social Media is Dead. Long Live Snapchat!”

“Separating the social from the media.”Axios op-ed, 2017

A small design core

A tiny team reviews hundreds of ideas a week with him in the room.

2018Decisions ran hub-and-spoke, with him at the center.

2026A design team of 9 to 12 people with no titles.

“The best way to have a good idea is to have lots of ideas.”Diary of a CEO, 2025

Kind, smart, creative

Kindness first, defined as honest support rather than niceness.

2022Named the three values in his layoff memo.

2025New hires pitch an idea on their first day.

“Kindness is the essential ingredient if you want to have a creative culture.”Diary of a CEO, via Fortune, 2025

Software gets copied

Features will be cloned, so invest in ecosystems, distribution and hardware.

2016Instagram copied Stories; he later called borrowing part of innovating.

2026Named Lens Studio, creators and Specs as the moats.

“Software is easy to copy.”Lenny’s Podcast, 2026

Pay for the long bet

Cut hard elsewhere to protect augmented reality.

Aug 2022Cut about 20% of staff and narrowed to three priorities.

Apr 2026Cut about 16% and over $500M of costs while protecting Specs.

“Until we get computing off a screen and into the world.”Fast Company, 2026

Evolution

How the thinking changed

YearsEraWhat they believedWhat changed it

2011–2013

Stanford and Picaboo

Disappearing photos are fun and freeing.

Renamed Snapchat; teens adopted it; a co-founder dispute later settled for $157.5M.

2013–2015

Hypergrowth

A generational company beats a $3B exit.

Stories and Discover extended ephemerality from chat to broadcasting.

2016–2017

Camera company and IPO

Snap is hardware, not only an app.

Spectacles, Instagram cloning Stories, a non-voting IPO.

2017–2019

Redesign crisis

Separate friends from media.

Daily users fell to 186M. His 2018 memo admitted the redesign was rushed.

2020–2023

Recovery

Augmented reality and subscriptions diversify revenue.

Snapchat+ launched; an ad slowdown forced a 20% cut.

2024–2026

Efficiency and Specs

AI enables small squads; computing moves to the face.

The 14-year letter, a failed Perplexity deal, a 16% cut, Specs launched.

Scale

The numbers behind the posture

493M

Daily active users, Q2 2026

25M

Snapchat+ subscribers, February 2026

~$3B

Facebook offer he turned down, 2013

$24B

Valuation at the 2017 IPO, with no votes for public shareholders

186M

Daily users at the low point after the 2018 redesign

~16%

Staff cut in April 2026 to fund the Specs bet

Tensions

Where the model strains

Kindness as the first value

Repeated deep layoffs and decisions centered on him

“Not social media”

An ad-funded business that pushes algorithmic content

Privacy and ephemerality

Cameras worn on the face

“Software is easy to copy”

Stories and Lenses, his proof of invention, were cloned

Long-horizon patience

Public shareholders without votes, and calls for him to step aside

For operators

What to take from it

Decide what your product opens to. It defines the product.

Protect a long bet with structure, not just conviction.

Assume features will be copied; invest in what can’t be.

If kindness is a value, show it most in how you cut.

Sources

Stanford talk via CBS SF (2013) · Forbes (2014) · TechCrunch (2014, 2016, 2022) · Snap S-1 (2017) · Fortune (2017, 2022, 2025) · Axios (2017) · Engadget (2018, 2026) · Entrepreneur (2024) · Diary of a CEO (2025) · Snap newsroom (2025, 2026) · Lenny’s Podcast (2026) · Fast Company and CNBC (2026) · Inc. (2026). Quotes shortened.

All profilesMelanie Perkins

Co-founder and CEO of Canva · 2007 to 2026

Melanie
Perkins

The Crazy-Big-Goal Setter

She started from a small pain, design software too hard to learn, proved the fix in one niche, and set a goal big enough to fund giving most of it away.

Drawn from Masters of Scale (2020), 20VC (2016), Guy Kawasaki (2020), Lenny’s Podcast (2025), her 2021 note to the Canva community, Canva’s “21 Questions” (2023), and interviews with Forbes, Fortune and Startup Daily.

The central plan

The two-step plan

Canva’s published plan since 2021. Step two is the reason for step one, and the founders have pledged most of their equity to it.

Swipe sideways to see the full diagram

Step oneBuild one of the world’s most valuable companiesAbout $4B in annualized revenue by the end of 2025265M monthly users, 31M payingProfitable since 2017

Step twoDo the most good we can30% of founder equity pledged in 2021, now over 80%$100M more to GiveDirectly in 2025A goal of basic needs met for everyone by 2050

Operating system

Eight principles

Start from a painful complexity

She watched students struggle with design software and fixed one narrow case first.

2007Fusion Books: school yearbooks designed online, started in her mother’s living room.

2013Launched Canva as the broad version of the same idea.

“They would take a whole semester just to learn the very basics.”Masters of Scale, 2020

Rejection is redirection

More than 100 investors said no. Each no rewrote the pitch.

2010–13About three years of trips from Perth to Silicon Valley.

2020Her pitch rule: show the problem before the solution.

“See each rejection as redirection.”Canva, 21 Questions, 2023 (shortened)

Wedge the door open

When a door moves an inch, push.

2011–12Learned to kitesurf to get time with investor Bill Tai.

2012Persisted with Cameron Adams after he declined; he became the third co-founder.

“Risk: serious damage; reward: start company.”Her note on kitesurfing, via Forbes, 2019

Crazy big goals

Set targets that look absurd, then work backward.

2019Said Canva had done 1% of what it could, at a $3.2B valuation.

2025Set a goal of basic needs met for everyone by 2050.

“Set crazy big goals and make them happen.”Startup Daily, 2024

Onboarding is the product

Remove fear at the first click; the free tier funnels to paid.

2013–14Users were afraid to touch buttons, so Canva added playful starter challenges.

2016–24Free, then Pro, Teams and Enterprise.

New users “felt confident and they got a little spark of joy.”Masters of Scale, 2020

Capital discipline

Profitable early, and slow to spend on what doesn’t pay.

2017Profitable, with 294,000 paying customers.

Aug 2026Slowed an AI rollout rather than ship at negative unit economics.

“We decided to slow the rollout while we rebuilt the architecture.”Fortune, 2026

Values you would pay for

A value is real only if you would give up revenue to keep it.

2021Published the two-step plan and pledged 30% of founder equity.

2025Gave $100M more to GiveDirectly.

“True values are things that you’d happily forego revenue to uphold.”Canva, 21 Questions, 2023

Integrate, don’t reinvent

Buy or partner for capability, then fold it into one simple product.

2024Acquired Affinity and Leonardo.ai; later made Affinity free.

2026Launched Canva AI 2.0 with outside model partners.

On Magic Studio: “the first all-in-one AI design platform on the market.”Fortune, 2023

Evolution

How the thinking changed

YearsEraWhat they believedWhat changed it

2007–2011

Perth, Fusion Books

Prove simple online design in one niche.

Yearbooks worked and gave her a platform to pitch from.

2011–2013

Outsider fundraising

Persistence and a better pitch will win.

Bill Tai’s network, Cameron Adams, seed funding, launch in 2013.

2013–2017

Product-led growth

Onboarding and templates are the moat.

Freemium conversion paid for growth; profitable by 2017.

2018–2021

Scale with purpose

Values are the operating system.

Valuation reached $40B; the two-step plan went public.

2022–2024

Suite and AI

A visual suite for teams, supercharged by AI.

Valuation cut to $26B, then Magic Studio, Affinity, Leonardo and Enterprise.

2025–2026

AI-native, pre-IPO

Integrated AI beats fragmented tools.

About $4B in annualized revenue, but AI costs cut 2026 growth guidance to 20%; an IPO now looks later.

Scale

The numbers behind the posture

100+

Investors who said no before the first round, 2010 to 2013

~$4B

Annualized revenue at the end of 2025

265M

Monthly users at the end of 2025

31M

Paying users, about 12% of monthly users

80%+

Share of the founders’ stake pledged to charity

$42B

Valuation in the August 2025 employee share sale

Tensions

Where the model strains

Profitable for years on near-zero marginal cost

AI inference costs forced a cut in growth guidance

Partner rather than reinvent

Margins now exposed to AI model vendors

Built for non-designers

Now owns a professional suite and sells to enterprises

Perth identity and patient “1%” framing

A Delaware parent company and IPO pressure

Giving away 80%+

The founders keep the stake, and control, until value is realized

For operators

What to take from it

Prove the idea in the smallest niche that feels the pain most.

Treat each rejection as feedback on the pitch, not the idea.

Write your values as things you’d pay to keep.

Know your marginal cost. AI can change it overnight.

Sources

Masters of Scale (2020) · 20VC (2016) · Guy Kawasaki (2020) · Lenny’s Podcast (2025) · Medium, “A note to the Canva community” (2021) · Canva, 21 Questions (2023) · canva.com/step-two · Forbes (2019) · Startup Daily (2024) · Fortune (2023, 2025, 2026) · Bloomberg (2025) · TechCrunch (2024) · SaaStr (2026) · Forbes Australia (2026). Quotes shortened.

All profilesMadam C.J. Walker

Founder of the Madam C.J. Walker Manufacturing Company · 1867 to 1919

Madam C.J.
Walker

The Distribution Builder

Born to formerly enslaved parents and orphaned at seven, she built a national company whose real product was economic independence: thousands of Black women trained to earn their own living.

Drawn from her own words (her 1912 National Negro Business League speech, letters, advertisements) and A’Lelia Bundles’ biography On Her Own Ground (2001), with the Smithsonian, National Archives, Library of Congress, Indiana Historical Society and Harvard Business School.

The central system

Distribution is people

She didn’t only sell hair products. She sold a way to earn: a method, training, a territory and a club to belong to. The agents were the channel and, in her eyes, the point.

Swipe sideways to see the full diagram

Product

Wonderful Hair Grower, launched in Denver, 1906.

→

Method

The Walker System: shampoo, grower, brushing, heated comb.

→

Training

Lelia College of Beauty Culture, from 1908.

→

Agents

About 1,000 by 1910; nearly 20,000 trained by 1917.

→

Clubs and causes

Walker Clubs, the 1917 convention, anti-lynching work.

One agent’s account in 1917, as reported by Walker’s biographer $5 a week as a servant → $250 a week as an agent

Operating system

Eight principles

Give yourself a start

Don’t wait for permission or an invitation.

1905Started her own product line in Denver with savings of about $1.25.

1912Ignored by Booker T. Washington at the National Negro Business League, she took the floor anyway.

“Don’t sit down and wait for the opportunities to come.”Quoted in Smithsonian Magazine

Distribution is people

The sales force was as much the product as the pomade: an income for Black women.

1910About 1,000 agents.

1917Nearly 20,000 women trained in the Walker System.

“Endeavoring to provide employment for hundreds of women of my race.”Quoted in Essence

Standardize, then train

A repeatable method taught at scale kept quality consistent.

1908Opened Lelia College of Beauty Culture in Pittsburgh.

1910Indianapolis headquarters combined factory, lab, salon and school.

“I have always held myself out as a hair culturist. I grow hair.”Dallas Express, 1919

The founder is the brand

Her name, face and story were on every tin.

1906Took the name Madam C.J. Walker; her own face went on the packaging.

1912Told her rags-to-riches story from the stage as part of the pitch.

She said the formula came to her in a dream, and told the story often.National Park Service (paraphrase)

Own the ground

Production and property of her own, not dependence on others.

1910Built a factory and headquarters in Indianapolis.

1918Moved into Villa Lewaro, designed by Black architect Vertner Tandy.

“I have built my own factory on my own ground.”1912 National Negro Business League speech

Turn a sales force into a movement

Clubs, conventions and prizes for both sales and charity.

1917First national convention of Walker agents, among the first national meetings of American businesswomen.

1917Delegates telegrammed President Wilson protesting lynching.

“I want my agents to feel that their first duty is to humanity.”Chicago Defender, 1918

Profit serves the mission

Wealth as leverage for her community.

1911Gave $1,000 to the Black YMCA in Indianapolis.

1919Gave $5,000 to the NAACP anti-lynching fund; her will left two-thirds of future profits to charity.

“There is no royal, flower-strewn road to success.”Quoted in Essence

Go wide early

She took the model abroad and into the courts within a decade of founding.

1913Sales trip to Cuba, Jamaica, Haiti, Costa Rica and Panama.

1915Sued an Indianapolis theater for discrimination.

The league that ignored her in 1912 gave her a speaking slot in 1913.Bundles, On Her Own Ground (paraphrase)

Evolution

How the thinking changed

YearsEraWhat they believedWhat changed it

1867–1887

Survival

Work is survival.

Orphaned at seven, married at fourteen, widowed around twenty; moved to St. Louis.

1888–1904

The washtub

Earn enough to educate her daughter.

A laundress on about $1.50 a day. Hair loss turned her own problem into a market.

1903–1905

Apprentice agent

Learn the product and the agent model.

Sold Annie Turnbo Malone’s products, then moved to Denver and broke away.

1905–1910

Independent founder

Build her own brand and train others to sell it.

Wonderful Hair Grower, the Walker name, Lelia College.

1910–1916

Scale and legitimacy

Business success earns respect.

The Indianapolis factory, the 1912 speech, Caribbean expansion, a discrimination lawsuit.

1916–1919

Power and politics

Agents should lead civic causes; wealth is leverage.

New York, the 1917 convention, a White House delegation on lynching, her will.

Scale

The numbers behind the posture

~20,000

Women trained as Walker agents by 1917

$1.50

Her daily pay as a laundress in the 1890s

$1,366

Revenue in her first year of business

$500K+

Sales in her final year, around 1918 to 1919

200+

Delegates at the first national convention of Walker agents, 1917

$600K+

Estate at her death in 1919; estimates run to $1M

Tensions

Where the model strains

Scalp health and hair growth

Critics tie her to Eurocentric beauty standards; she avoided the word “straightener”

Built her own brand

Built on Annie Turnbo Malone’s product and sales model

“First self-made female millionaire”

Her biographer estimates about $600,000, and she denied the title herself

Headline agent counts

1,000, 20,000 and 40,000 all get cited; trained, active and employed differ

The self-made story

Church networks, her husband’s advertising skill and Malone’s playbook all helped

For operators

What to take from it

Design your distribution as an income opportunity for the people doing it.

Write the method down and teach it. Consistency scales; charisma doesn’t.

Give your sales force belonging: clubs, prizes, a cause.

If you aren’t invited to speak, take the floor.

Sources

Madam C.J. Walker, 1912 National Negro Business League speech (Library of Congress records) · A’Lelia Bundles, On Her Own Ground (2001) · Smithsonian Magazine · National Archives · Library of Congress · Indiana Historical Society · Encyclopedia of Indianapolis · AAIHS · HBS Baker Library · National Park Service · National Women’s History Museum · Essence · Salon (2018) · Chicago Defender (1918). Quotes shortened.

All profilesJeff Bezos

Founder of Amazon and Blue Origin · 1994 to 2026

Jeff
Bezos

The Day 1 Inventor

He runs every year as Day 1: start from the customer, decide fast on anything reversible, write the thinking down and let failures grow with the company. In 2026 he is testing the method again as a co-CEO.

Drawn from his shareholder letters (1997 to 2020), Invent and Wander (2020), his 2011 shareholder meeting remarks, his 2021 email to employees, Lex Fridman (2023), DealBook (2024) and his Fortune interview (2026).

The central idea

It is always Day 1

Day 1 is a large company that still behaves like a startup. Day 2 is what follows when it stops. His 2016 letter named four defenses, and he closed his 2021 handover email with “It remains Day 1.”

Swipe sideways to see the full diagram

Day 1Four habits that keep a company in Day 1Customer obsessionA skeptical view of proxiesThe eager adoption of external trendsHigh-velocity decision making

Day 2StasisFollowed by irrelevanceFollowed by excruciating, painful declineFollowed by death

Wording from his 2016 shareholder letter, published April 2017. “Day 1” first appeared in the 1997 letter, which Amazon still reprints in its annual report, most recently for 2025.

Operating system

Eight principles

Customers, not competitors

Customers are never satisfied for long, so they are a better guide to what to build than rivals are.

Feb 2005Launched Prime: $79 a year for free two-day shipping.

2026Told Fortune he wants Amazon to take pride in serving customers, not in being big.

“A lot of companies will tell you they’re customer-obsessed, but they’re really competitor-obsessed.”Fortune, 2026

It’s all about the long term

Judge the business by long-run cash flows and market leadership, not by the next quarter’s accounting.

1997First letter: sales up 838% to $147.8M, with a pledge to invest boldly for market leadership.

2003First full-year profit, $35M, nine years after founding.

“We will make bold rather than timid investment decisions.”1997 shareholder letter

One-way and two-way doors

Most decisions are reversible and should be made fast. Only the irreversible few deserve slow, heavy process.

Apr 2016The 2015 letter split decisions into Type 1, one-way doors, and Type 2, two-way doors.

Apr 2017Set a rule of thumb: decide with about 70% of the information you wish you had.

“Many decisions are reversible, two-way doors. Those decisions can use a light-weight process.”2016 shareholder letter

Disagree and commit

With conviction but no consensus, ask for commitment rather than agreement, and give it yourself.

Apr 2017Wrote that he approved an Amazon Studios original he doubted, and committed to it.

Apr 2017Told teams to escalate real misalignment at once instead of arguing to exhaustion.

“Look, I know we disagree on this but will you gamble with me on it? Disagree and commit?”2016 shareholder letter

Failure has to scale

A big company that runs only small experiments stops inventing. One large winner pays for many losers.

2014The Fire Phone led to a $170M loss; its lessons and developers went into Echo and Alexa.

May 2026A New Glenn rocket exploded in a pad test. He said Blue Origin would rebuild and fly again.

“Amazon will be experimenting at the right scale… if we occasionally have multibillion-dollar failures.”2018 shareholder letter

Memos, not slides

A written narrative exposes weak thinking that slides hide, and gets everyone to the same starting point.

Apr 2018Wrote that Amazon does not do PowerPoint; meetings open with a six-page memo read in silence.

Dec 2023Said the most senior person in the room should speak last, so others say what they think.

“My perfect meeting starts with a crisp document… and a messy meeting.”Lex Fridman, 2023

High standards are teachable

People learn high standards by exposure, one domain at a time, and need a realistic view of the work involved.

c. 2002Amazon’s leadership principles were first written down, about eight years after founding.

Apr 2018Used a handstand lesson to make the point: unrealistic beliefs about scope kill high standards.

“I believe high standards are teachable.”2017 shareholder letter

Willing to be misunderstood

Inventing at scale means looking wrong to outsiders for years, so he does not manage to the reaction.

2011Told shareholders it took three tries to make third-party selling work.

Dec 2024At DealBook, called ending the Post’s presidential endorsements “far from cowardly.”

“We are willing to be misunderstood for long periods of time.”2011 shareholder meeting, via GeekWire

Evolution

How the thinking changed

YearsEraWhat they believedWhat changed it

1986–1994

Wall Street

He would regret not trying the internet more than leaving a good job.

Left D. E. Shaw, drove to Seattle and started Amazon in a rented Bellevue garage with $300,000 from his parents.

1994–2001

Get Big Fast

Scale first. Market leadership makes the economics work.

The dot-com crash. His 2000 letter opened “Ouch.” The first quarterly profit came in the last quarter of 2001.

2002–2010

Platform builder

Open Amazon’s own machinery to sellers, developers and readers.

Prime in 2005, AWS in 2006 and Kindle in 2007 turned one retailer into several businesses.

2011–2020

Day 1 at scale

A giant must decide fast and fail big to stay inventive.

The Fire Phone failed and Echo worked. He bought the Post for $250M in 2013 and Whole Foods for $13.7B in 2017.

2021–2024

Executive chair

Amazon can run on its principles while he works on space, climate and the Post.

Flew on New Shepard in July 2021. In 2024 he ended the Post’s endorsements and more than 200,000 subscribers cancelled.

2025–2026

Operator again

AI for engineering and manufacturing is worth an operating job.

Co-CEO of Prometheus, valued near $41B. A New Glenn exploded in May 2026 and Blue Origin sought its first outside investors.

Scale

The numbers behind the posture

838%

Amazon sales growth in 1997, to $147.8M, the year of the Day 1 letter

$717B

Amazon net sales in 2025; No. 1 on the 2026 Fortune Global 500

58%

Third-party share of physical goods sales in 2018, up from 3% in 1999

~$30B

His own money put into Blue Origin from 2000 to September 2026

$41B

Valuation of Prometheus after its $12B round, June 2026

1/3

Washington Post staff cut in February 2026; bought for $250M in 2013

Tensions

Where the model strains

Pledged in 2021 to make Amazon “Earth’s Best Employer”

About 30,000 corporate jobs cut between October 2025 and January 2026

Called ending Post endorsements “a principled decision”

200,000+ subscribers cancelled; a third of the staff was cut in February 2026

Customer obsession as the first rule

The FTC and 17 states sued in 2023, alleging an illegally maintained monopoly

Wanted to build a company that would outlast him

Still executive chair; Amazon paid his Blue Origin about $1.8B in 2025

For operators

What to take from it

Sort decisions by reversibility. Go fast on two-way doors, slow on one-way doors.

Decide at about 70% of the information, then correct bad calls quickly.

Swap slides for a written memo, read it together, and let the senior person speak last.

Let the size of your failed experiments grow with the company.

Sources

Amazon shareholder letters (1997, 2014, 2016, 2017, 2018, 2020) · Jeff Bezos, Invent and Wander (2020) · Amazon, email to employees (February 2021) · GeekWire (2011, 2024, 2025, 2026) · Lex Fridman Podcast (2023) via Fortune · DealBook Summit (2024) via GeekWire · Fortune (2019, 2023, 2024, 2026) · NPR and AP (2026) · Amazon 2025 annual report, Q4 2025 results and 2026 proxy · FTC (2023) · Benzinga and The Motley Fool (2026). Quotes shortened.

All profilesMarc Benioff

Co-founder, chair and CEO of Salesforce · 1999 to 2026

Marc
Benioff

The Alignment Evangelist

He runs Salesforce from a short plan that ranks values ahead of methods, sells by picking a bigger enemy, and built giving into the company at founding. Since 2025 his own conduct has tested the values he ranked first.

Drawn from Behind the Cloud (2009), Trailblazer (2019), his V2MOM essay on the Salesforce blog, his CES keynote (2020), Fortune’s Leadership Next (2024), The Logan Bartlett Show (2025), earnings calls (2025 and 2026), the Dreamforce 2026 keynote and Salesforce’s 2026 filings.

The central method

V2MOM: five questions in order

Vision, values, methods, obstacles, measures. He wrote the first one on an envelope in April 1999 and it became the business plan. Every employee now drafts their own each year.

Swipe sideways to see the full diagram

1 · What do you wantVisionRapidly create a world-class Internet company/site for Sales Force Automation.

→

2 · What is importantValuesWorld-class organization. Time to market. Functional. Usability (Amazon quality). Value-added partnerships.

→

3 · How do you get itMethodsHire the team. Finalize the product. Develop to beta and production. Build partnerships. Build a launch plan. Exit: IPO or acquisition.

→

4 · What is in the wayObstaclesDevelopers. A product manager and business development person.

→

5 · How do you knowMeasuresPrototype is state-of-the-art. High-quality functional system. Partnerships online. Regarded as leader. “We are all rich.”

Each year every department and every employee drafts their own V2MOM, and all of them are published on the company’s internal network.

Box contents are from Salesforce’s first V2MOM, dated 12 April 1999 and written on an American Express envelope, as reproduced in Behind the Cloud (2009). Lists shortened.

Operating system

Eight principles

Five questions, in order

Vision, values, methods, obstacles, measures: one format from the company plan down to each employee.

Apr 1999Wrote the first V2MOM on an American Express envelope; it became the business plan.

Jun 2004Parker Harris had kept the envelope and gave it to him framed on the day of the IPO.

“Success depends on constant communication and complete alignment.”V2MOM essay, Salesforce blog

Trust is the first value

He ranks trust above growth and says a company that does not will pay for it.

2018Set up an Office of Ethical and Humane Use after a dispute over a border agency contract.

Sep 2026Told Dreamforce that customer data is not retained by, or used to train, outside models.

“Trust has to be your highest value in your company.”Quoted in Fortune, 2019

Give 1% from the start

The 1-1-1 model commits 1% of equity, product and employee time to philanthropy from founding.

2014Co-founded Pledge 1% so other companies could adopt the model.

Apr 2026Proxy reported nearly $1B in grants and over 10M volunteer hours to date.

“We put the seed of giving back at the beginning.”CES keynote, January 2020

Always go after Goliath

Define the company against the market leader so customers see an underdog with a cause.

Feb 2000Staged a mock “End of Software” protest, with a fake TV crew, at Siebel’s user conference.

Dec 2024Said Microsoft’s Copilot had been “a disappointment” to many customers.

“If there is no Goliath in your industry, go after the status quo.”Behind the Cloud, 2009

Keep a beginner’s mind

Let go of what worked before and ask what is happening now; he credits Zen practice for it.

1996Took a sabbatical from Oracle to Hawaii and India and came back with the idea for Salesforce.

Sep 2024Launched Agentforce at Dreamforce and recast the company around AI agents.

“In your beginner’s mind, you have every possibility, but in your expert’s mind, you have few.”Fortune Leadership Next, 2024

Business as a platform for change

A chief executive answers to employees, customers and communities, and should use that leverage.

Mar 2015Cancelled Salesforce travel to Indiana over a religious freedom law; it was revised in a week.

Nov 2018Funded San Francisco’s Proposition C, a business tax for homeless services; it passed.

“All stakeholders have to matter for us, not just shareholders.”CES keynote, January 2020

Buy the missing layer

When the platform lacks a layer, such as integration, analytics, chat or data, he acquires it.

2018–21Bought MuleSoft for $6.5B, Tableau for $15.7B and Slack for $27.7B.

Nov 2025Closed the roughly $8B Informatica deal to feed Agentforce clean data.

“You have to get your data right to get your AI right.”Salesforce press release, November 2025

Be your own first customer

He sells AI agents as digital labor and offers Salesforce’s own headcount as the proof.

Sep 2025Said support staff had gone from 9,000 to about 5,000 as agents took over cases.

May 2026Said engineering had held near 15,000 people for two years; new hiring was mostly in sales.

“I’ve reduced it from 9,000 heads to about 5,000, because I need less heads.”The Logan Bartlett Show, 2025

Evolution

How the thinking changed

YearsEraWhat they believedWhat changed it

1979–1999

Coder to Oracle VP

Enterprise software is licensed, installed and sold from the top.

Thirteen years at Oracle under Larry Ellison, then a 1996 sabbatical in Hawaii and India.

1999–2004

The End of Software

Business software should be rented over the internet.

Founded with three engineers in March 1999; the June 2004 IPO raised $110M.

2004–2012

Platform and cloud

A sales app can become a platform that others build on.

AppExchange in 2005, $1B in annual revenue by 2009, Dreamforce as the yearly rally.

2013–2021

Acquirer and activist

Values create value; buy what the platform lacks.

ExactTarget, MuleSoft, Tableau and Slack; Indiana in 2015; Time magazine and Proposition C in 2018.

2022–2024

Reset under pressure

A co-CEO could share the load while growth continued.

Bret Taylor left, five activist funds took stakes, and he cut about 10% of staff in January 2023.

2024–2026

Agentic pivot

AI agents are digital labor, and Salesforce will sell it.

Agentforce passed $1.5B in recurring revenue by July 2026; the shares were still down over 25% by August.

Scale

The numbers behind the posture

$41.5B

Revenue in fiscal 2026, the year to January 2026, up 10%

$1.5B+

Agentforce annual recurring revenue, July 2026, up over 240%

$27.7B

Slack, closed July 2021, his largest acquisition

83,334

Employees at January 31, 2026

4,000

Support roles cut in 2025 as AI agents took over cases

~$1B

Community grants since 1999 under the 1-1-1 model, per the 2026 proxy

Tensions

Where the model strains

Trust and equality as the stated core values

Joked that ICE was tracking foreign staff, Feb 2026; over 1,300 signed a protest letter

The activist CEO who fought Indiana’s 2015 law

Backed Trump sending the National Guard to San Francisco in Oct 2025, then apologized

“Humans still have an incredible role in business”

4,000 support roles cut in 2025 and no engineers hired in fiscal 2026

“Not just shareholders”

10% of staff cut in 2023 under activist pressure; a $50B buyback authorized in 2026

For operators

What to take from it

Write the plan as five answers, in order, and make every team write theirs.

Rank your values. The order is the decision when two of them collide.

Set aside the 1% at founding; it is far harder to add later.

If you sell a product as labor, expect to be judged on your own headcount.

Sources

Marc Benioff and Carlye Adler, Behind the Cloud (2009) · Marc Benioff and Monica Langley, Trailblazer (2019) · Salesforce blog, “How to Create Alignment Within Your Company” · Trailhead · LEADERS (2012) · HuffPost (2015) · CES keynote via GeekWire (2020) · Fortune (2019, 2023, 2024, 2026) · The Logan Bartlett Show via NBC News (2025) · San Francisco Standard (2025) · diginomica (2025, 2026) · CNN and KQED (2025) · Salesforce press releases, 10-K and proxy (2025, 2026) · Dreamforce 2026 keynote transcript · Salesforce Ben, KRON4 and Gazetteer SF (2026) · Inc. (2026) · The Motley Fool (2026). Quotes shortened.

All profilesRodney Sacks

Co-founder and chairman of Monster Beverage · 1990 to 2026

Rodney
Sacks

The Asset-Light Lawyer

A corporate lawyer who bought a failing juice brand, he kept what earns a premium: the name, the can and the next flavour. Others make the drinks, Coca-Cola’s bottlers move them, and he defends the rest on the record.

Drawn from his Beverage Industry interview (2011), Senate testimony (2013), the Coca-Cola partnership announcements (2014 and 2015), Monster earnings calls (2020 to 2024), SEC filings (2025 and 2026) and his 2025 SA Jewish Report profile.

The central model

Own the brand, rent the rest

Monster designs, prices and markets the drink. Co-packers make most of it and Coca-Cola’s bottlers deliver it, so the money goes into the brand instead of plants and trucks.

Swipe sideways to see the full diagram

Monster keepsBrand, formula, priceMarketing, new products and the flavours, which its own AFF plants make.

→

Others makeCo-packersThird-party bottlers and contract packers made the majority of its energy drinks in 2025.

→

Others moveCoca-Cola bottlersTwo of them were about 25% of 2025 net sales. Coca-Cola owns about 20.9% of Monster.

→

Others sellRetailersHis 2011 case for shelf space: energy drinks ring up higher than soft drinks.

$8.29BNet sales, 2025

55.8%Gross margin, 2025

$2.42BOperating income, 2025

6,891Employees, end of 2025

Figures from Monster’s results and annual report for the year to 31 December 2025. The two bottlers are Coca-Cola Europacific Partners (15%) and Coca-Cola Consolidated (10%). Coca-Cola’s stake is as of February 2026.

Operating system

Eight principles

Keep the fixed costs outside

Let co-packers and bottlers own the plants and trucks; keep the brand, the formula and the cash.

1992Bought the Hansen’s business, with 12 employees and about $17.5M in sales, for $14.5M.

2025Still outsourced most energy drink production: 6,891 staff on $8.29B of net sales.

“Outsource the manufacturing process for the majority of our finished goods energy drink products.”Monster annual report for 2025 (company filing)

Sell an image, give more can

Compete on identity and on value at once: a bigger drink than the leader for about the same price.

Apr 2002Launched Monster Energy in a 16-ounce black can, about twice Red Bull’s size for a similar price.

2005Monster ranked second only to Red Bull; Hansen’s nine-month sales nearly doubled to $251M.

“We really were selling an image.”SA Jewish Report, 2025

Borrow the best distribution

Trade equity and side brands for reach that would take decades to build alone.

2006–08Signed Anheuser-Busch wholesalers in 2006, then added Coca-Cola bottlers two years later.

Jun 2015Closed the Coca-Cola deal: about 16.7% of Monster for $2.15B net, plus its bottling system.

“We gain enhanced access to The Coca-Cola Company’s distribution system.”Coca-Cola and Monster release, Aug 2014

One brand, one meaning

A brand stands for one thing. Stretch it too far and the meaning thins.

2011Sold tea as Peace Tea in 99-cent, 24-ounce cans rather than put the Monster name on it.

Jun 2015Handed Hansen’s sodas, juices and Peace Tea to Coca-Cola and took its energy brands.

“What we realized is that brands are not everything to everyone.”Beverage Industry, 2011

Expect to be wrong, cheaply

Launch often, change fast and walk away before much money is sunk.

1996–97Tried an energy formula in a smoothie in 1996, then Hansen’s Energy in 8.3-ounce cans in 1997.

2022–25Bought CANarchy’s craft brewers for $330M; reported about $130M of alcohol impairments in February 2025.

“Knowing that you’re going to make mistakes and you may have to change.”Beverage Industry, 2011

Listen to the younger guys

Test products and packaging on young staff and athletes in their own setting, not on focus panels.

2002 onDesigned Monster for masculine appeal in a black can and backed it with action-sports athletes.

2011Said packaging and promotions were run past junior staff; Monster had 12.3M Facebook fans.

“We listen to the younger guys.”Beverage Industry, 2011

Answer every attack on the record

A lawyer’s reflex: contest claims in testimony and in court instead of letting them stand.

Jul 2013Testified to a US Senate committee on energy drink marketing and caffeine.

Sep 2022Monster won a $293M false advertising verdict against Bang’s maker, then bought Bang in 2023.

“The Company does not market Monster to children, and has never done so.”Senate written testimony, July 2013

A small team of opposites

He credits a few unlike people who push aggressive ideas and pull each other back.

1990Formed Hansen Natural with Hilton Schlosberg; Mark Hall later ran the Monster unit.

Jun 2025Handed the CEO job to Schlosberg, co-CEO since 2021 and his partner since 1990.

“We’re very, very different people.”Beverage Industry, 2011

Evolution

How the thinking changed

YearsEraWhat they believedWhat changed it

1970–1989

Johannesburg lawyer

A deal is only as good as the integrity of the other party.

Made partner young at Werksmans, then left South Africa for California in August 1989 to build something for his family.

1990–1996

Hansen’s turnaround

A tired natural soda brand can be revived with new products and no factories.

Paid $14.5M for Hansen’s in 1992, lost money through 1995, then made $357,000 in 1996 on a new smoothie line.

1997–2001

Functional drinks

Drinks that do something can command a premium price.

Hansen’s Energy launched in 1997. Sales reached about $80M by 2001, but Red Bull still owned the category.

2002–2011

Monster

Energy is not a fad; sell an image in a bigger can.

Monster launched in April 2002 and was second to Red Bull by 2005. The company took the Monster name in January 2012.

2012–2020

Scrutiny and Coca-Cola

Defend the product in public; go global through someone else’s system.

Wrongful death suits and a 2013 Senate hearing, then the Coca-Cola deal, closed in June 2015.

2021–2026

Handover

Succession can be gradual: share the job, then keep only the parts you enjoy.

Shared the CEO job from 2021 and left it in June 2025. Chairman through 31 December 2026, then a non-employee director.

Scale

The numbers behind the posture

$14.5M

Paid for the Hansen’s business in 1992, when it had 12 employees

$8.29B

Monster net sales in 2025, up 10.7% on 2024

6,891

Employees worldwide at 31 December 2025

$2.15B

Net cash from Coca-Cola for about 16.7% of Monster, closed June 2015

7.5%

Sacks’s stake in Monster, per the 2026 proxy statement

Kiplinger’s best stock of the century: up 155,103%, 2000 to June 2025

Tensions

Where the model strains

“Does not market Monster to children”

In 2015 Monster settled wrongful death suits, one over a 14-year-old, on undisclosed terms

Asset-light: others make and move the product

Coca-Cola owns 20.9% of the stock; two of its bottlers were 25% of 2025 sales

Brands are “not everything to everyone”

Bought brewers for $330M in 2022; reported about $130M of alcohol impairments in 2025

“We don’t have any politics”

Five women’s suits, reported in 2018, alleged harassment and bias; Monster denied them

For operators

What to take from it

Own what customers pay a premium for. Rent the plants and the trucks.

Give a partner equity if it buys reach you could not build in a decade.

Launch cheaply enough that walking away from a product costs little.

Know what your brand means, and stop extending it where the meaning ends.

Sources

Beverage Industry, Executive of the Year interview (2011) · US Senate Commerce Committee written testimony (July 2013) · Coca-Cola and Monster press releases (August 2014, June 2015) · Monster earnings calls (2020, 2024) · Monster 2025 results, 10-K for 2025, 2026 proxy statement and March 2025 8-K · SA Jewish Report (2025) · Los Angeles Times (2005) · BusinessWeek (2005) · FundingUniverse · BeverageDaily (2012, 2014) · BevNET (2015, 2024) · HuffPost (2018) · Food Dive (2025) · Kiplinger (2025) · Orange County Business Journal (2026) · Beverage Digest. Quotes shortened.

All profilesTodd Vasos

CEO of Dollar General · 2015 to 2022, 2023 to 2026

Todd
Vasos

The Small-Box Operator

He puts small stores in small towns for a shopper who runs short before month end, and competes on distance and price. When execution slipped, he came back and returned the company to store basics.

Drawn from Dollar General earnings calls (2019 to 2026), his annual report letters (2024 to 2026), The Wall Street Journal (2017), Chain Store Age (2025), company releases (2022, 2023, 2026) and reporting by CNBC, Supermarket News and Retail Dive.

The central model

The small box, five miles away

A small, low-cost store in a small town, stocked for a shopper who buys what she needs today. He repeated it until about three in four Americans lived within five miles of one.

Swipe sideways to see the full diagram

WhereSmall townsAbout 80% of stores are in towns of 20,000 people or fewer.

→

FormatSmall boxAbout 7,500 square feet of selling space on average.

→

TripSmall basketConsumables were 82% of net sales in fiscal 2025.

→

PriceSmall pricesMore than 2,000 items priced at $1 or less.

→

Result21,148 storesWithin five miles of about 75% of the US population.

Town share, store size and consumables share from the fiscal 2025 annual report (year ended January 30, 2026). Store count at July 31, 2026. Price count and five-mile reach from management remarks, August and September 2026.

Operating system

Eight principles

Know one customer cold

He plans around households earning under $35,000, about 60% of sales in 2024.

Dec 2017Told the Journal she buys one bottle of ketchup on the day the last one runs out.

Aug 2024Reported that the three weakest sales weeks of the quarter were the last week of each month.

“No one knows our core customers better than we do.”Chain Store Age, 2025

Go where the big boxes are not

He builds in towns of 20,000 people or fewer, about 80% of the chain by fiscal 2025.

2015–22Added about 7,000 stores in his first term as CEO.

Mar 2025Said he still saw about 12,000 more sites for stores in the United States.

“The economy is continuing to create more of our core customer.”The Wall Street Journal, 2017

Price for the cash in hand

Price points are set for a shopper who counts the total before she reaches the register.

Dec 2017The Journal reported shelf prices marked in nickel increments.

Aug 2026Value Valley held more than 600 rotating $1 items; its comparable sales rose 16% in the quarter.

“They don’t want to be embarrassed when they get up to the register.”The Wall Street Journal, 2017

Own the supply chain

Distribute fresh and frozen food yourself to lower cost and keep shelves in stock.

2019Launched DG Fresh self-distribution, with a first facility in Pottsville, Pennsylvania.

Mar 2026Fresh produce had reached more than 7,000 stores.

“DG Fresh will eventually allow us to control our own destiny.”Q1 2019 earnings call, via Modern Retail

Back to basics

On his return he added no new strategy. He put hours, stock and people back into stores.

Dec 2023Put $150M into store labor hours for 2023 and moved smart-team hours to store teams.

2024Removed self-checkout from about 12,000 stores to keep an employee at the front.

“We are getting back to the basics here at Dollar General across all levels.”Q3 2023 earnings call

Stabilize the store manager

He treats store manager turnover as the first number to fix in a store that is slipping.

Dec 2023Told investors store manager turnover was higher than he wanted.

Jun 2025Reported lower year-to-date turnover at every level of retail operations.

“When we stabilize the store manager position, the entire store and team benefit.”Q3 2023 earnings call

Fix the fleet you have

Open fewer stores, remodel the old ones on a cycle and close those that cannot work.

Mar 2025Announced the closure of 96 Dollar General and 45 pOpshelf stores after a portfolio review.

2025Remodeled 4,254 stores under Project Renovate and Project Elevate; opened 589, down from 987 in 2023.

“We expect to touch approximately 20% of our store base annually.”Q1 2025 earnings call

The store is the delivery hub

He treats 20,000 small-town stores as a same-day delivery network.

2025Same-day delivery reached about 18,000 stores; over 80% of orders arrived within an hour.

Aug 2026Said more than 1 million new customers came first through delivery, then shopped in store.

“There is no one out there today that can deliver to small town rural America within an hour.”Q4 2024 earnings call

Evolution

How the thinking changed

YearsEraWhat they believedWhat changed it

1980s–2008

Drugstore operator

Retail is learned from the store floor up.

Started as an Eckerd assistant manager in Wilmington, North Carolina; rose to chief operating officer of Longs Drug Stores.

2008–2015

Merchant and COO

Assortment and price for a low-income shopper make the box work.

Joined as chief merchandising officer at about 8,400 stores, became chief operating officer, then CEO in June 2015.

2015–2019

Store-count machine

Rural America has room for thousands more small stores.

Passed 14,000 stores in 2017 and 15,500 in 2019, opening close to 1,000 a year.

2019–2022

New engines

The model can stretch to fresh food, self-checkout, pOpshelf and Mexico.

Launched DG Fresh, Fast Track and pOpshelf; fiscal 2020 sales reached $33.7B; he retired in November 2022.

2023–2024

Return to basics

Execution had slipped. The fix is labor, in-stocks and store standards.

Recalled in October 2023 after Jeff Owen was removed; shares fell 32% in one day in August 2024.

2025–2026

Repair and handover

Fewer openings, more remodels, margin back to 6% to 7% by 2028.

Fiscal 2025 EPS rose 34%; in March 2026 JJ Fleeman of Ahold Delhaize USA was named CEO from January 1, 2027.

Scale

The numbers behind the posture

21,148

Stores at July 31, 2026, in 48 states and Mexico

~7,000

Stores added in his first term as CEO, 2015 to 2022

$42.7B

Net sales in fiscal 2025, up 5.2%

~80%

Share of stores in towns of 20,000 people or fewer, fiscal 2025

−32%

One-day share price fall on August 29, 2024, after a guidance cut

$12M

Settlement with OSHA in July 2024 over store safety hazards

Tensions

Where the model strains

A mission of “Serving Others”

OSHA put the chain in its severe violator program; it paid $12M in 2024

Value for shoppers who count every dollar

Pennsylvania stores failed over 40% of price accuracy checks, 2019 to 2023

Back to basics removed self-checkout and paused pOpshelf

Both launched in his first term, in 2019 and 2020

Said in 2022 he was confident in his successor

The board removed Jeff Owen within a year and called Vasos back

For operators

What to take from it

Pick one customer and build the store around how she actually shops.

When results slip, check store basics before you add a new strategy.

Track store manager turnover. Fix it before you fix anything else.

Be willing to reverse your own launches when the numbers say so.

Sources

Dollar General earnings calls (Q1 2019, Q3 2023, Q1, Q2 and Q4 2024, Q1 2025, Q2 2026) · Annual report letters (2024 to 2026) · Form 10-K (fiscal 2025) · Dollar General press releases (2022, 2023, 2026) · The Wall Street Journal (2017) · Chain Store Age (2025) · Modern Retail (2019) · CNBC (2021, 2023, 2024) · Supermarket News (2023, 2024, 2026) · Retail Dive and Grocery Dive (2022, 2024, 2026) · Retail TouchPoints (2024) · The Shelby Report (2026) · WilmingtonBiz (2015) · KeyCorp director biography · Pennsylvania Attorney General (2025). Quotes shortened.

All profilesSundar Pichai

CEO of Google and Alphabet · 2004 to 2026

Sundar
Pichai

The Quiet Platform Builder

He names a platform shift early, funds it for a decade and owns every layer, from chips to products. He hears everyone out, then asks for commitment. The calm that lets him wait is what critics call slowness.

Drawn from his 2016 founders’ letter, Stanford GSB (2022), his January 2023 staff memo, Lex Fridman (2025), All-In (2025), the BBC (2025), Cheeky Pint (2026), TIME (2026), the I/O 2026 keynote and Alphabet earnings calls (2024 to 2026).

The central idea

AI-first, built as a full stack

He declared Google AI-first in 2016, then spent a decade owning each layer, from chips to products. Every layer was a bet placed years before it paid, and each one now feeds the next.

Swipe sideways to see the full diagram

Layer 1 · since 2015Custom chipsTPUs, announced in 2016 and now in their eighth generation.

→

Layer 2 · merged 2023One model teamBrain and DeepMind combined. Gemini 3 shipped in November 2025.

→

Layer 3 · billions of usersProductsGemini app at 950M monthly users. AI Mode in Search above 1B.

→

Layer 4 · sold to othersCloud$24.8B revenue in Q2 2026, up 82%, with a $514B backlog.

The loop closes with cash. Revenue goes back into layer 1 as capital spending: the 2026 guidance, up from $91.4B spent in 2025.$195B to $205B

Figures from Alphabet’s Q2 2026 results (July 2026), the I/O keynote (May 2026) and Cloud Next (April 2026). Pichai declared Google AI-first in April 2016.

Operating system

Eight principles

Call the shift early

Name the next platform years before it shows up in revenue, and reorganize around it.

Apr 2016His first founders’ letter as CEO said computing would move from mobile first to AI first.

May 2026Opened I/O by marking ten years since the AI-first pivot, with the Gemini app at 900M monthly users.

“We will move from mobile first to an AI first world.”2016 founders’ letter

Own the full stack

Chips, data centers, models and products are built to work together, so each gain compounds.

2015–16Put custom TPU chips into Google’s data centers and announced them at I/O in May 2016.

Apr 2026Launched eighth-generation TPUs; said just over half of 2026 ML compute spend would go to Cloud.

“Our AI investments and full-stack approach are driving performance across our business.”Q1 2026 earnings call

Bets take a decade

Judge a project by its value five to ten years out, and keep funding it while nobody is watching.

2008–09By his account, Chrome had 1% market share a year after launch; it held about 71% in December 2025.

2009–26Waymo began as a 2009 project and passed 500,000 driverless rides a week in early 2026.

“We work through these things, whether there’s attention from the outside or not.”All-In, 2025

Hear everyone, then break the tie

Bill Campbell asked him weekly which ties he had broken. He listens at length, then decides.

2021Fifteen current and former executives told the New York Times his decisions came too slowly.

Apr 2023Merged Google Brain and DeepMind into one unit, after DeepMind had pushed for independence for years.

“You make a clear decision and you just ask people to commit.”Lex Fridman, 2025

Tune out noise, watch signals

He treats most outside criticism as noise and judges progress by the trajectory he sees inside.

Feb 2023A factual error in the Bard demo cut about $100B from Alphabet’s market value in a day.

Jan 2026Alphabet passed $4T in market value, the fourth company to do so, after Gemini 3 and an Apple deal.

“I’m good at tuning out noise. I do watch out for signals.”Lex Fridman, 2025

Disrupt Search yourself

He put AI answers inside Search instead of protecting the results page that funds the company.

May 2024Rolled out AI Overviews to US Search, changing the page that earns most of Alphabet’s revenue.

2025–26AI Mode passed 1B monthly users a year after its US rollout; Search revenue grew 17% in Q2 2026.

“The dilemma only exists if you treat it as a dilemma.”All-In, 2025

Underinvesting is the bigger risk

When demand bends upward he would rather own spare capacity than miss the shift.

Nov 2025Told the BBC the AI boom had “elements of irrationality” and that no company would be immune.

Jul 2026Raised 2026 capex guidance to $195B to $205B, from $91.4B spent in 2025; the shares fell on the news.

“The risk of under-investing is dramatically greater than the risk of over-investing.”Q2 2024 earnings call

Scale without adding layers

Since 2023 he has paired record spending on AI with cuts to headcount and management.

Jan 2023Cut about 12,000 jobs, 6% of staff, and wrote that he took full responsibility.

Dec 2024Told staff Google had cut manager, director and vice-president roles by 10%.

“We hired for a different economic reality than the one we face today.”Memo to employees, January 2023

Evolution

How the thinking changed

YearsEraWhat they believedWhat changed it

1972–2004

Chennai to California

Access to technology changes what a family can do.

A five-year wait for a rotary phone; then IIT Kharagpur, Stanford, Wharton and McKinsey. Joined Google in 2004.

2004–2013

Toolbar and Chrome

Google needs its own route to users.

Chrome shipped in September 2008 after Eric Schmidt had resisted building a browser for years.

2013–2015

Android and product

One calm operator can align rival product groups.

Took over Android in March 2013 and product and engineering in October 2014; named Google CEO in August 2015.

2015–2019

AI-first CEO

Computing moves from mobile first to AI first.

Staff protests over Project Maven led to the 2018 AI principles (weapons pledge dropped in 2025); he took over Alphabet.

2019–2023

Alphabet and ChatGPT

Google had the pieces and could release AI carefully.

ChatGPT arrived in November 2022. He cut 12,000 jobs, shipped Bard with an error and merged Brain with DeepMind.

2024–2026

Full-stack payoff

Owning chips, models and products wins the shift.

Lost two monopoly cases but kept Chrome and the ad exchange; Gemini 3, a $4T valuation, capex near $200B.

Scale

The numbers behind the posture

$205B

Top of 2026 capex guidance, set July 2026; $91.4B was spent in 2025

950M

Gemini app monthly active users, Q2 2026

$514B

Google Cloud backlog, Q2 2026

$4T

Alphabet market value, first reached January 2026

12,000

Jobs cut in January 2023, about 6% of the workforce

$692M

Maximum value of his three-year pay package, granted March 2026

Tensions

Where the model strains

Hears everyone out, then asks people to commit

In 2021, 15 current and former executives told the NYT he decided too slowly

AI-first since 2016, with the Transformer built in-house

OpenAI shipped the chatbot first; he says Google was behind on frontier models

Argues Google wins because its products are better

US courts found illegal monopolies in search (2024) and ad tech (2025)

Took “full responsibility” for the 2023 layoffs

Rolling cuts in record 2026 quarters; 4,500 staff petitioned him; pay of up to $692M

For operators

What to take from it

Name the platform shift early, then fund it for a decade before it pays.

Own the layers your product depends on. Integration compounds.

Hear everyone out, then break the tie and ask for commitment.

Trust your internal trajectory, but count slow shipping as a real cost.

Sources

2016 founders’ letter via Entrepreneur · Larry Page, “G is for Google” (2015) · Stanford GSB (2022) · Lex Fridman #471 (2025) · All-In (2025) · Cheeky Pint (2026) · TIME (2026) · BBC (2025) · Alphabet earnings calls and releases (2024 to 2026) · Google I/O and Cloud Next keynotes (2026) · New York Times via Fortune (2021) · NPR (2023, 2025) · Business Insider via TechRadar (2024) · Al Jazeera (2025) · Fortune (2026) · Business Today (2026) · The Motley Fool (2026) · AdExchanger (2026) · Courthouse News (2026) · StatCounter via Wikipedia (2025). Quotes shortened.

All profilesMary Barra

Chair and CEO of General Motors · 2014 to 2026

Mary
Barra

The Accountable Insider

A GM lifer who treats bad news as something to face the minute it is known. She apologized for the ignition switch, left markets and projects that did not pay, and wrote down EV plans instead of defending them.

Drawn from her 2014 employee town hall and congressional testimony, Wharton People Analytics (2018), Stanford GSB (2024), Fortune (2024, 2026), Decoder (2025), Axios (2026), and GM shareholder letters and earnings calls (2025 to 2026).

The central reflex

Solve it the minute you know

Her rule since the 2014 ignition-switch crisis: face a problem as soon as it is known, say it in public, pay to fix it, and keep the memory so the next one surfaces sooner.

Swipe sideways to see the full diagram

Step oneFace itFebruary 2014: began recalling 2.6 million cars, weeks after becoming CEO. The defect had an 11-year history inside GM.

→

Step twoSay itApril 2014, to Congress: “I am deeply sorry.” June 2014: told employees worldwide what the outside inquiry found.

→

Step threeFix it15 people dismissed, Speak Up for Safety launched, a victims’ fund that accepted 124 death claims.

→

Step fourKeep it“I never want to put this behind us.” The failure stays in GM’s memory on purpose.

The memory feeds back to step one. Later problems she named and closed: Opel sold in 2017, Cruise robotaxi funding ended in December 2024, $10.9B of EV charges booked from late 2025 to mid 2026.

Figures: 2.6 million cars recalled (February to March 2014); 15 dismissed (June 2014); 124 deaths accepted by the compensation fund (final tally, 2015). Quotes from her April 2014 testimony and 5 June 2014 town hall. EV charges are GM’s reported totals through the second quarter of 2026.

Operating system

Eight principles

Solve it the minute you know

Problems grow. Raise them while they are still small enough to fix.

Feb 2014Began recalling 2.6 million cars weeks into the job, for a switch defect with an 11-year history.

Dec 2024Took more than $5B of charges to restructure GM’s loss-making China joint ventures.

“The best time to solve a problem is the minute you know you have one.”Fortune Most Powerful Women, 2024

Never put it behind you

Keep a failure in the company’s memory so that it keeps changing behavior.

Jun 2014Told a global town hall the inquiry had found “a pattern of incompetence and neglect.”

Jun 2014Dismissed 15 people and disciplined five more over the ignition switch.

“I never want to put this behind us.”Employee town hall, 5 June 2014

Apologize, then be transparent

Say sorry first, bring in an outside investigator, and publish what is found.

Apr 2014Opened her House testimony with an apology to the victims and their families.

2014–15Set up a compensation fund run by Kenneth Feinberg; it accepted 124 death claims.

“Today’s GM will do the right thing.”House testimony, April 2014

Behaviors set the culture

Culture is what people do each day, so give managers judgment, not rulebooks.

2009As HR chief, replaced a 10-page dress code with two words: “Dress appropriately.”

2014Launched Speak Up for Safety: staying silent about a defect makes you “part of the problem.”

“I strongly believe behaviors set your culture.”Stanford GSB, 2024 (shortened)

Returns over volume

If a market has no path to profit, leave it and hand the cash to shareholders.

Mar 2017Sold Opel and Vauxhall to PSA for about $2.2B, ending nearly 90 years in Europe.

2023–26Authorized $28B of buybacks in four programs; the share count fell about 35% by January 2026.

“We are transforming our business, establishing GM as a more focused and disciplined company.”GM statement, May 2017

Destination fixed, timing flexible

She keeps the all-electric end point and moves the dates when policy or demand shifts.

Jan 2021Set an aspiration to end tailpipe emissions from new light-duty vehicles by 2035.

2025–26Booked about $10.9B of EV charges and switched the Orion plant to gasoline trucks.

“We’re mainly investing and continuing to work on EVs because we think that’s the end game.”Automotive Press Association, January 2026

Admit the missed forecast

She names her own wrong calls, then moves the bet to something GM can ship.

Dec 2024Ended Cruise robotaxi funding after more than $10B invested since 2016.

2026Super Cruise passed 1 billion miles; eyes-off highway driving is planned for 2028.

“I started predicting when we would have level 4, or true autonomy, and I got it wrong.”Axios, January 2026

Own the software in the car

GM builds its own in-car system and sells subscriptions on top of it.

Oct 2025Said gasoline models would also lose Apple CarPlay at major refreshes, as GM’s EVs already had.

2025OnStar reached a record 12 million subscribers; Super Cruise passed 620,000.

“We needed to have a great system in the vehicle that allowed people to have one system.”Decoder, October 2025

Evolution

How the thinking changed

YearsEraWhat they believedWhat changed it

1980–2009

Co-op to plant manager

Hard work and a united team fix most problems.

Joined at 18 inspecting fenders at Pontiac, took a Stanford MBA on a GM fellowship, ran the Detroit-Hamtramck plant from 2003.

2009–2013

HR and product chief

Fewer rules and clearer accountability make better cars.

After the 2009 bankruptcy she cut the dress code to two words, then ran global product development from 2011.

2014–2015

Ignition-switch reset

Face the failure in public and never let the company forget it.

2.6 million cars recalled, 15 people dismissed, 124 death claims accepted, a $900M federal settlement in 2015.

2016–2020

Shrink to profit

Returns beat volume; leave markets with no path to profit.

Chair from 2016. Sold Opel, left India, South Africa, Australia and Thailand, cut 14,700 jobs in 2018, a 40-day strike in 2019.

2021–2024

All-electric bet

An all-electric lineup by 2035, robotaxis and software would remake GM.

Missed the 1 million EV capacity goal, took $5B of China charges, began $10B of buybacks, ended Cruise robotaxis in 2024.

2025–2026

Reset to trucks

EVs are still the end game, but gasoline trucks and US plants pay for now.

$10.9B of EV charges, $3.1B of 2025 tariffs, guidance raised twice in 2026. No successor named; she says she has more to give.

Scale

The numbers behind the posture

124

Death claims accepted by the ignition-switch fund, final tally 2015

$10.9B

EV realignment charges, second half of 2025 through June 2026

$28B

Share buybacks authorized, November 2023 to January 2026

17.2%

US market share in 2025, GM’s highest since 2015

$3.1B

Tariff costs in 2025; EBIT-adjusted was still $12.7B

Years at GM in 2026; she joined as a co-op student in 1980

Tensions

Where the model strains

An all-electric aspiration for 2035; EVs as “the end game”

An EV plant switched to gasoline trucks and nearly $1B for new V8 engines

“Win their hearts and minds” on the factory floor

14,700 job cuts announced in 2018 and a 40-day UAW strike in 2019

Solve problems the minute you know

Predicted driverless cars by 2019; Cruise ended in 2024 after $10B+

Capital discipline while investing in the future

$28B of buybacks authorized as $10.9B of EV investment was written off

For operators

What to take from it

Raise the problem the minute you see it. It will not get smaller.

After a failure, publish the findings and keep the memory on purpose.

Leave markets and projects with no path to profit, even after billions spent.

Keep the destination, and say out loud when the dates change.

Sources

Barra town hall remarks via Automotive World (June 2014) · Congressional testimony via WardsAuto and Fortune (2014) · CBS News (2015) · Knowledge@Wharton and Quartz (2018) · CNBC (2019) · Stanford GSB (2024) · Fortune (2023, 2024, 2026) · Decoder (2025) · Axios (2026) · InsideEVs (2024, 2026) · GM shareholder letters (Q3 2025 to Q2 2026) · GM Q1 2026 earnings call · GM Q2 2026 results · ABC News (2024) · GM Authority (2026). Quotes shortened.

All profilesWarren Buffett

Chairman emeritus, Berkshire Hathaway · 1965 to 2026

Warren
Buffett

The Patient Compounder

He treated insurance float as cheap, long-lived capital, put it into businesses he understood, kept every dollar they earned and let it compound for sixty years. He wrote to shareholders as partners.

Drawn from his shareholder letters (1988 to 2024), An Owner’s Manual (1999 text), “The Superinvestors of Graham-and-Doddsville” (1984), his 2014 “Past, Present and Future” letter, the May 2025 annual meeting, his November 2025 Thanksgiving letter and his September 2026 statement.

The central loop

Compound the float

Insurance customers pay first and claim later. Berkshire invests the money in between, keeps what its businesses earn, and uses the larger balance sheet to write more insurance.

Swipe sideways to see the full diagram

Step oneCollect premiumsInsurance customers pay up front. Claims are paid later, sometimes decades later.

→

Step twoHold the floatMoney Berkshire holds but does not own. About $20M in 1967, $176B at the end of 2025.

→

Step threeBuy good businessesWhole companies and stocks he understands, bought at a fair price and kept.

→

Step fourKeep the earningsProfits go back to Omaha to be reinvested rather than paid out.

A bigger balance sheet can write more insurance, which feeds back to step one.

The result: compounded annual gain in Berkshire’s per-share market value, 1965 to 2025, against 10.5% for the S&P 500 with dividends.19.7% a year

Float figures are from Berkshire’s 2004 shareholder letter (1967) and its 2025 annual report (31 December 2025). Return figures are from the performance table in the 2025 annual report.

Operating system

Eight principles

Stay inside the circle

Judge only businesses you understand, and be honest about where that understanding ends.

1999Sat out the internet boom. Berkshire shares fell 19.9% while the S&P 500 returned 21%.

2016Berkshire began buying Apple with about $1 billion of stock; it became the largest holding.

“The size of that circle is not very important; knowing its boundaries, however, is vital.”1996 shareholder letter

Wonderful business, fair price

Charlie Munger moved him from cheap, mediocre companies to great ones worth paying for.

1972Paid $25 million for See’s Candies, which had earned $1.35 billion before tax by 2007.

1985Closed Berkshire’s textile mills after twenty years of owning a cheap, poor business.

“It’s far better to buy a wonderful company at a fair price than a fair company at a wonderful price.”1989 shareholder letter

Float is the engine

Premiums arrive before claims are paid, and Berkshire invests the money in between.

1967Bought National Indemnity for $8.6 million. It came with about $20 million of float.

1996Paid $2.3 billion for the 49% of GEICO that Berkshire did not already own.

“Float, which is money that doesn’t belong to us but that we temporarily hold.”2004 shareholder letter

The holding period is forever

Buy to keep. Compounding needs time and as few interruptions as possible.

1988–94Bought 400 million Coca-Cola shares for $1.3 billion and still held every one in 2022.

2022Coca-Cola’s yearly dividend to Berkshire reached $704 million, up from $75 million in 1994.

“Our favorite holding period is forever.”1988 shareholder letter

Greedy when others are fearful

Hold spare cash so that other people’s panics become buying opportunities.

Sep 2008Bought $5 billion of Goldman Sachs preferred stock eight days after Lehman Brothers failed.

2025Ended his last year as CEO with about $373 billion in cash and Treasury bills.

“Be fearful when others are greedy and greedy only when others are fearful.”2004 shareholder letter

Delegate almost to abdication

Buy good managers along with the business, then leave them alone.

1999Berkshire had about 45,000 employees, 12 of them at headquarters.

2024–25About 388,000 employees worldwide; the Omaha head office had 27 at the end of 2024.

“We delegate almost to the point of abdication.”An Owner’s Manual, 1999 text

Reputation before profit

A firm can afford to lose money. It cannot afford to lose its good name.

Aug 1991Became interim chairman of Salomon after its Treasury bid scandal, which cost the firm $290 million.

Sep 2026Made his son Howard chairman to guard “culture and values” while Greg Abel runs the company.

Lose money for the firm and I will be understanding. Lose reputation and I will be ruthless.Salomon testimony, 1991 (paraphrase)

Put mistakes in writing

He reports his own errors to shareholders by name, with the cost attached.

1993Paid $433 million in Berkshire stock for Dexter Shoe, whose value soon went to zero.

2014Wrote that folding National Indemnity into Berkshire cost his partners about $100 billion.

“To date, Dexter is the worst deal that I’ve made.”2007 shareholder letter

Evolution

How the thinking changed

YearsEraWhat they believedWhat changed it

1950–1956

Graham’s student

Buy what is statistically cheap, whatever the business.

Studied under Benjamin Graham at Columbia, worked at Graham-Newman from 1954, then went home to Omaha.

1956–1969

The partnership

Cigar butts: one free puff is enough.

Returned 29.5% a year against 7.4% for the Dow, took control of textile maker Berkshire in 1965 and closed the fund in 1969.

1970–1985

Float and franchises

Pay up for quality and fund it with insurance float.

National Indemnity’s float and See’s Candies (1972) worked; the textile mills did not and closed in 1985.

1986–1999

Permanent holdings

A few great businesses, held for good.

Coca-Cola, GEICO and General Re. Salomon nearly failed in 1991; in 1999 the stock fell 19.9% as the S&P rose 21%.

2000–2018

Elephant hunting

Size is an anchor: buy whole companies and accept lower returns.

Bought the rest of BNSF for $26B in 2010 and Precision Castparts for $32.1B in 2016, then built the Apple stake.

2019–2026

Cash and handover

Few prices make sense, and the culture has to outlast him.

Munger died in 2023. Greg Abel became CEO in January 2026; in September Buffett became chairman emeritus.

Scale

The numbers behind the posture

19.7%

Yearly gain in Berkshire’s share price, 1965 to 2025; S&P 500: 10.5%

$176B

Insurance float at the end of 2025, up from about $20 million in 1967

$373B

Cash and Treasury bills at the end of 2025, his last day as CEO

$8.6M

Paid for National Indemnity in 1967, the first source of float

$60B+

Berkshire shares given away between 2006 and mid-2026

2034

Year by which he plans to have given away his last shares (July 2026)

Tensions

Where the model strains

“Our favorite holding period is forever”

Sold every airline stock in April 2020 and nearly half the Apple stake in 2024

Berkshire “will never prefer” cash to owning good businesses

It sold more stock than it bought for 14 straight quarters to March 2026

Hands-off trust in his managers

In 2011 he first defended David Sokol’s Lubrizol trades, then called them inexcusable

Pledged more than 99% of his wealth to philanthropy

His July 2026 gifts went only to family foundations, and his son now chairs Berkshire

For operators

What to take from it

Secure cheap, durable capital before you go looking for places to put it.

Define the edge of what you understand, and say no to everything beyond it.

Pay a fair price for a great business rather than a low price for a poor one.

Report your mistakes to your owners in writing, with the cost attached.

Sources

Berkshire Hathaway shareholder letters (1988, 1989, 1996, 2000, 2004, 2007, 2022, 2023, 2024) · An Owner’s Manual (1999 text) · “Berkshire: Past, Present and Future” (2014) · “The Superinvestors of Graham-and-Doddsville” (1984) · Thanksgiving letter (November 2025) · Berkshire 2025 annual report (2026) · Berkshire press release (18 September 2026) · CNBC (2019, 2025) · AP (2025, 2026) · Fortune (2025, 2026) · Reuters via The National (2026) · Fox Business (2026) · 24/7 Wall St. (2026). Quotes shortened; paraphrases marked.

All profilesSheryl Sandberg

COO of Facebook and Meta · 2008 to 2022

Sheryl
Sandberg

The Founder’s Operator

She took a founder’s product and built the business beside it: hire ahead of growth, sell attention as advertising, trade blunt feedback every week. The machine she built later made her answer for what it did with people’s data.

Drawn from Lean In (2013), Option B (2017), her TEDWomen talk (2010), commencement speeches at Harvard Business School (2012), UC Berkeley (2016) and Brandeis (2026), Masters of Scale (2017), her Senate testimony (2018) and her 2022 departure post.

The central build

The business beside the product

Zuckerberg kept the product. She built what paid for it: an advertising business that took revenue from $272M in 2008 to $117.9B in 2021, her last full year as COO.

Swipe sideways to see the full diagram

She joins as COO2008

$272M

IPO year2012

$5.09B

Five years on2017

$40.7B

Her last full year2021

$117.9B

She steps down2022

$116.6B

Facebook and Meta annual revenue from company filings, bars to scale. Sandberg was COO from March 2008 to 1 August 2022. 2022 was the company’s first annual revenue decline.

Operating system

Eight principles

Get on the rocket ship

Choose growth over title. Fast-growing companies create roles that do not exist yet.

2001Joined Google as its first business unit general manager, on Eric Schmidt’s advice.

May 2026Told Brandeis graduates to drop the ten-year plan for a short-term direction and a long-run dream.

“If you’re offered a seat on a rocket ship, don’t ask what seat. Just get on.”Eric Schmidt, quoted in her HBS speech, 2012

Give the product a business

A popular product still needs a model. Hers was advertising that builds awareness and demand.

2008Facebook’s leaders settled on advertising as the business within months of her arrival.

2021Revenue reached $117.9B, up from $272M in 2008.

“You can think about it how I think about it, which is, it’s awareness building.”Fortune, April 2008

Hire for aptitude, ahead of need

New markets have no veterans. Hire able people early and plan several steps ahead.

2001–08Grew Google’s online sales and operations team from 4 people to 4,000.

Jun 2022Zuckerberg credited her with hiring great people and forging Meta’s management culture.

“I did not go look for people with online ad sales experience.”Masters of Scale, 2017

Make it safe to tell you

Leaders hear the truth only if they ask for it and make it safe to give.

2000sAt Google her interviews had become a hiring bottleneck, and nobody said so until she asked.

2008 onMet Zuckerberg first thing on Monday and last thing on Friday to trade feedback.

“I realized I had to make it safe to speak up when I’m messing up.”Masters of Scale, 2017

Commit with the founder

The operator’s job is to back the founder’s big call in full, then make it pay.

Mar 2008Left Google to join a 23-year-old founder at a 550-person Facebook.

2012Backed his mobile-first order after the IPO, at a cost to near-term desktop ad revenue.

“You really can’t be fired, and you’re the only person who can fire me. If you’re in, I’m in.”To Zuckerberg, via Masters of Scale, 2017

Sit at the table

She argued that women are held back and also hold themselves back, and told them to push.

Dec 2010TEDWomen talk: sit at the table, make your partner a real partner, don’t leave before you leave.

Mar 2013Published Lean In and founded LeanIn.Org; the book sold 4.2M copies in five years.

“We hold ourselves back in ways both big and small.”Lean In, 2013 (shortened)

Resilience is built

Grief taught her that recovery is a skill that people and organizations can practise.

May 2015Her husband, Dave Goldberg, died suddenly on 1 May.

Apr 2017Published Option B with Adam Grant, on facing adversity and building resilience.

“Like a muscle, you can build it up, draw on it when you need it.”UC Berkeley commencement, 2016

Own it on the record

When the company fails, say so in public and take the blame for your part.

Sep 2018Told the Senate Intelligence Committee Facebook had been too slow on Russian interference.

Nov 2018After the Definers report, wrote that she oversaw communications and took full responsibility.

“We were too slow to spot this and too slow to act.”Senate testimony, September 2018

Evolution

How the thinking changed

YearsEraWhat they believedWhat changed it

1991–2001

Washington

Work that matters is done in public service, beside a mentor.

Research assistant to Larry Summers at the World Bank, then his chief of staff at the US Treasury from 1996 to 2001.

2001–2008

Google

Join the fast-growing company and the role will grow with it.

Built online sales and operations from 4 people to 4,000, then met Mark Zuckerberg at a Christmas party in late 2007.

2008–2012

Building the business

A popular site needs a business model, and the model is advertising.

Revenue rose from $272M in 2008 to $5.09B in 2012; the May 2012 IPO valued Facebook at $104B.

2012–2017

Mobile and Lean In

Go mobile-first at work; tell women to sit at the table.

Lean In sold millions, Dave Goldberg died in May 2015, Option B followed in 2017, and revenue reached $40.7B.

2018–2022

Crisis and exit

Admit Facebook was too slow, then fix it from the inside.

Cambridge Analytica, the Definers report and a $5B FTC penalty in 2019; she stepped down as COO on 1 August 2022.

2022–2026

Investor and advocate

Back founders, fund Lean In, and name the backsliding on women at work.

Left Meta’s board in May 2024, was sanctioned in 2025, joined Nscale’s board and named a new Lean In CEO in March 2026.

Scale

The numbers behind the posture

$272M

Facebook revenue in 2008, the year she joined as COO

$118B

Meta revenue in 2021, her last full year, nearly all of it advertising

4,000

Her Google sales and operations team by 2008, up from 4 people in 2001

4.2M

Copies of Lean In sold in the five years after March 2013

$5B

FTC privacy penalty, July 2019, after the Cambridge Analytica scandal

$190M

Cambridge Analytica shareholder settlement, approved April 2026

Tensions

Where the model strains

Advertising that gives small businesses the reach of large ones

The data practices behind it drew a $5B FTC privacy penalty in 2019

Took “full responsibility” for Facebook’s communications team

Its firm Definers tied critics to George Soros; she said she did not remember the firm

Told the Senate in 2018 that Facebook had been too slow to act

Sanctioned in 2025 for deleting emails in the Cambridge Analytica shareholder case

Lean In: women should sit at the table and push

Michelle Obama said it “doesn’t work all the time”; bell hooks called it faux feminism

For operators

What to take from it

Join the fastest-growing thing you can and let the role form around you.

Decide what business you are in before you scale the team that sells it.

Book feedback with your boss every week, and make it safe for your team to correct you.

If you take responsibility in public, keep the records that let others check it.

Sources

Sheryl Sandberg, Lean In (2013) and Option B (2017) · TEDWomen (2010) · HBS Class Day speech (2012) · UC Berkeley commencement via NBC News (2016) · Masters of Scale (2017) · Senate testimony via Fortune (2018) · Facebook Newsroom (2018) · Facebook posts via CNN (2022) and Fortune (2024) · Fortune (2008, 2025, 2026) · Axios (2022) · The New York Times and Vice (2018) · AP (2025) · Bloomberg and Bloomberg Law (2025, 2026) · TechCrunch (2022, 2026) · Forbes (2026). Quotes shortened.

All profilesJack Ma

Co-founder of Alibaba and Ant Group · 1999 to 2026

Jack
Ma

The Schoolteacher Evangelist

He ranked customers first, employees second and shareholders third, designed Alibaba to outlive him, and backed small firms against banks and giants. One speech about regulators in 2020 tested every part of that.

Drawn from his 1999 apartment speech, Stanford talk (2011), IPO prospectus letter (2014), Davos sessions (2015, 2018), succession letter (2018), Bund Summit speech (2020) and staff memos and remarks (2023 to 2026), several in translation.

The central idea

Customers, employees, then shareholders

His stated order of whom Alibaba serves, written into the 2014 prospectus so investors were told before they bought. The Alipay transfer and his 996 remarks show where it strained.

Swipe sideways to see the full diagram

FirstCustomersTaobao launched free for sellers in 2003 and passed eBay’s share in China by 2005Eight customers, not executives, rang the opening bell at the 2014 IPOHis 2023 instruction to managers: go back to Taobao, go back to users

SecondEmployees18 founders in a Hangzhou apartment in 1999The Alibaba Partnership, formalised in 2010, nominates a majority of the board18 partners in 2026; he sits on its five-member committee

ThirdShareholders$25B raised in New York in September 2014His prospectus letter admitted the order could be “a bit hard to understand”Alipay left Alibaba in 2011; Yahoo and SoftBank said they were not told first

The order is as stated in his letter in Alibaba’s 2014 IPO prospectus. Events dated 1999 to 2023; partner count from Alibaba’s investor relations site, 2026.

Operating system

Eight principles

Customers first, shareholders last

Serve customers, then staff; shareholder returns follow from those two, and he told investors so.

Sep 2014Eight customers, not executives, rang the New York opening bell at the $25B IPO.

May 2023Told Taobao and Tmall managers to go back to Taobao, to users and to the internet.

“Customers first, employees second, and shareholders third.”IPO prospectus letter, 2014

A company for 102 years

Aim to span three centuries, which means building a structure that does not need its founder.

Jul 2010Formalised the Alibaba Partnership, which nominates a majority of the board.

Sep 2019Handed the chair to Daniel Zhang on Alibaba’s 20th anniversary, his own 55th birthday.

“No company can rely solely on its founders. Of all people, I should know that.”Succession letter, September 2018

Fight in the river

Pick the ground where local knowledge beats a larger rival’s scale.

1999Told co-founders in his apartment that the real competitors were in Silicon Valley.

2003–05Launched Taobao free for sellers; it passed eBay’s market share in China by 2005.

“We are the crocodiles in the Yangtze river. Never fight in the ocean.”Stanford China 2.0, 2011

Hire people smarter than you

He never wrote code. He saw his job as finding better people and keeping them together.

2011At Stanford he called himself “a blind man riding on the back of blind tigers.”

2013–15Gave up the CEO post in May 2013; Daniel Zhang held it from 2015.

“My job is to make sure smart people are working together.”Davos, 2018

Rejection is the curriculum

He treats failure as ordinary and tells the stories against himself.

1982–84Failed the university entrance exam twice and passed at the third attempt.

2001By his account, US venture capitalists turned down his request for $5M.

“Twenty-three people were accepted. I was the only guy who wasn’t.”On a KFC job, Davos with Charlie Rose, 2015

Change the banks

Use data and payments to give small firms the credit that collateral-based banks refuse.

Oct 2013Yu’e Bao became China’s largest money market fund; he had vowed to change the banks.

Oct 2020Told the Bund Summit banks had a “pawnshop mentality.” Ant’s IPO was halted on 3 November.

“We cannot use yesterday’s way to manage the future.”Bund Summit, 2020 (translation by Kevin Xu)

Name the mistake, then reform

Say in public what went wrong and change before the market forces it.

Nov 2023Congratulated rival Pinduoduo on the staff forum as its market value neared Alibaba’s.

Apr 2024Wrote that Alibaba was operating on its own “diseases of a big company.”

“It is not the making of mistakes that is fearsome; everyone makes them.”Staff memo, April 2024 (translation)

Teach what machines can’t

In an AI era, schools and companies should build values, creativity and judgment, not recall.

Apr 2025Told Alibaba Cloud staff that AI should understand humans, not be made to resemble them.

Mar 2026Took Alibaba’s and Ant’s chiefs to Yungu School to discuss teaching in the AI era.

“We cannot teach our kids to compete with the machines.”Davos, 2018

Evolution

How the thinking changed

YearsEraWhat they believedWhat changed it

1988–1998

Teacher, China Pages

The internet could put China’s small firms in front of the world.

Taught English, launched China Pages in 1995, saw it pass to a state telecom firm, then worked at a trade ministry.

1999–2004

Apartment years

Chinese teams can match Silicon Valley by serving small traders.

18 founders pooled 500,000 yuan; Goldman Sachs and SoftBank invested; Taobao (2003) and Alipay (2004) followed.

2005–2013

Scale and control

Customers first, with insiders keeping control of the mission.

Yahoo paid $1B for 40% in 2005. The 2011 Alipay transfer angered Yahoo and SoftBank. He left the CEO job in May 2013.

2014–2019

IPO and evangelism

A 102-year company has to outlast its founder.

A $25B New York listing, Davos stages, a backlash over his 996 remarks, then the chair passed to Daniel Zhang.

2020–2022

Speech and silence

Old financial rules were holding back credit for small firms.

His Bund speech on 24 October 2020, Ant’s halted IPO, an 18.2B yuan fine on Alibaba, and a retreat to Tokyo.

2023–2026

Quiet return

AI restarts the race, so reform before you are forced to.

Gave up Ant control, wrote staff memos, attended Xi’s 2025 symposium, and went back to campus with no executive title.

Scale

The numbers behind the posture

102

Years Alibaba was meant to last, a goal he set at its 1999 founding

$25B

Raised in the September 2014 New York IPO, a record at the time

$37B

Ant Group IPO halted on 3 November 2020, days after his Bund speech

¥18.2B

Antitrust fine on Alibaba, April 2021 (about $2.8B)

6.2%

His Ant voting rights after January 2023, down from more than 50%

$76.5M

Alibaba shares he reportedly bought in August 2026 (HK$600M)

Tensions

Where the model strains

Customers first, shareholders third

In 2011 Alipay left Alibaba; Yahoo and SoftBank said they were not told first

“Be in love with the government, but don’t marry them”

A 2020 speech on regulators preceded Ant’s halted IPO and his loss of control there

Employees rank second, above shareholders

Called the 996 schedule “a huge blessing” in 2019 and drew a public backlash

“No company can rely solely on its founders”

Retired in 2019; by 2025 reported to be steering AI and subsidy decisions

For operators

What to take from it

Write down who you serve first, and tell investors before they buy.

Build the succession structure a decade before you need it.

Fight a bigger rival on the ground you know best.

If a regulator can stop your business, weigh every public word about it.

Sources

Crocodile in the Yangtze (2012 film, 1999 speech) · Stanford China 2.0 transcript (2011) · Alibaba IPO prospectus letter and staff memo (2014) · Davos with Charlie Rose (2015) · World Economic Forum (2018) · Succession letter (2018) · WGBH (2014) · Reuters (2019) · Al Jazeera (2020) · Bund Summit speech, translated by Kevin Xu (2020) · Bloomberg and Fortune (2023, 2024, 2025) · Pekingnology (2024) · SCMP (2023, 2025, 2026) · Caixin (2025, 2026) · Alibaba investor relations (2026) · Alizila (2013) · Duncan Clark, Alibaba (2016). Several quotes are English translations from Chinese. Quotes shortened.

All profilesSoichiro Honda

Founder of Honda Motor · 1906 to 1991

Soichiro
Honda

The Shop-Floor Engineer

He made the machines and left money, sales and organisation to Takeo Fujisawa. He set targets at world level, treated a failure as information, and used racing to force his engineers to learn faster.

Drawn from Honda Motor’s official history in English translation, which quotes his Honda Monthly essay (1953), company newsletters (1959 to 1969), his 1965 Grand Prix statement, a 1983 television interview and his 1991 farewell message.

The central partnership

The maker and the seller

Honda’s own history puts it in one line: Honda did the making and Fujisawa did the selling. For 24 years one ran engineering, the other ran money, sales and organisation, and they left together.

Swipe sideways to see the full diagram

Soichiro Honda · President, 1948 to 1973The makingEngines, products and the factory floor1947: asked of every drawing what in it was new1961: first five places in the 125cc and 250cc Isle of Man TT races1972: the CVCC engine, with about 230 patent applications pending

Takeo Fujisawa · Managing director from 1949The sellingMoney, dealers and the organisation1952: wrote to bicycle shops and chose 13,000 as dealers1954: asked suppliers to accept 30% payments and secured Mitsubishi Bank’s support1959: chose the United States first; American Honda opened with $250,000

What each said of the other, in the company’s own account:“He’s got what I haven’t got.”

Roles, dates and figures from Honda Motor’s official history in English translation. The two met in August 1949 and stepped down together in 1973.

Operating system

Eight principles

Split the making from the selling

He kept to engineering and gave Fujisawa the money, the dealers and the running of the company.

Oct 1949Fujisawa joined as managing director; a month later he put up a quarter of the new capital.

1952Fujisawa wrote to bicycle shops across Japan and chose 13,000 as dealers for the Cub F-Type.

“He’s got what I haven’t got.”Both men, in Honda’s company history (translated)

Aim at the world, not Japan

He set world-level targets while the company was still a small workshop in Hamamatsu.

1951Told a job applicant that Honda would be the world’s biggest motorcycle maker.

1968Pushed for the CB750 Four after a trip to Switzerland showed bikes sized for Japan looked small.

“If we aren’t Number One in the world, we won’t be Number One in Japan.”Honda company history, D-Type chapter (translated)

Failure is information

A flop that shows you are on the wrong track is a gain, so do the thing you think is right.

1930sOnly 3 of the first 50 piston rings from his firm Tokai Seiki passed Toyota’s inspection.

Oct 1965After Honda’s first Formula One win he told staff to study wins as closely as losses.

“Just finding out that you’re on the wrong track is getting a lucky break.”Recalled by Kawashima, Honda history (translated)

Race to learn

He entered the hardest races so his engineers had to reach world standard in public.

Mar 1954Announced Honda would enter the Isle of Man TT, weeks before a cash crisis halted production.

1961Honda took the first five places in both the 125cc and 250cc TT races.

“We decided to choose the most difficult path to get there.”Press conference, October 1965 (translated)

Aim for 120% quality

People aiming at 100% miss by 1%, and that 1% is a wholly defective product to its buyer.

Mar 1953Set out the 120% standard in the Honda Monthly newsletter.

Apr 1954Apologised in person to Saitama factory workers for a carburettor fault in the Dream 4E.

“To eliminate the possibility of missing by 1%, we should aim for 120% quality.”Honda Monthly, March 1953 (translated)

Ask what is new

He judged a design by what was original in it and held that ideas, not capital, make a business.

1947Reviewed engineers’ drawings by asking what was new and what differed from other makes.

1972Introduced the CVCC low-emission engine; Toyota, Ford, Chrysler and Isuzu later licensed it.

“What part of this is new? What part is different from other makes?”Honda company history, 1947 (translated)

Do not wait for permission

He refused to let a ministry decide which companies would be allowed to build cars.

1963Launched the T360 truck and S500 sports car, Honda’s first production cars.

1983Recalled on television how he had “deluged” a ministry official with complaints over the bill.

“The government couldn’t tell me what to do.”TV interview, 1983, via Honda history (translated)

Step back when you are the block

When his own engineering conviction held the company up, he gave way, then handed over.

1969Chose to “stay as president” and let his engineers switch from air to water cooling.

1973Retired at 66 together with Fujisawa; Kiyoshi Kawashima became president.

“Do whatever you want, but make sure you take care of the water problems.”Honda company history, H1300 chapter (translated)

Evolution

How the thinking changed

YearsEraWhat they believedWhat changed it

1922–1945

Mechanic and supplier

Skilled hands and trial and error can solve any engineering problem.

His piston rings failed Toyota’s tests until 1939. Bombing and a 1945 earthquake wrecked his plants and he sold out to Toyota.

1946–1949

Hamamatsu workshop

Ideas, not capital, are the basis of a business.

Fitted 500 surplus generator engines to bicycles, built the A-Type, and incorporated Honda Motor in 1948 with 34 people.

1949–1954

Partnership and crisis

He makes, Fujisawa sells, and the target is number one in the world.

Dealers reached 13,000. In 1954 a carburettor fault and the failed Juno scooter left Honda unable to pay suppliers in full.

1954–1963

Racing and Super Cub

Race against the world’s best to learn; sell a machine anyone can ride.

Super Cub in 1958, American Honda and the first TT entry in 1959, world’s largest motorcycle maker, TT sweep in 1961.

1963–1973

Cars and conflict

Honda is free to make cars, and air cooling is the right way to build them.

The air-cooled H1300 sold poorly; his air-cooled F1 car crashed fatally in 1968. Engineers won water cooling and built the Civic.

1973–1991

Supreme adviser

The founders should leave together and let younger people run the company.

Retired with Fujisawa in 1973, toured dealers to thank them, became supreme adviser in 1983 and died on 5 August 1991.

Scale

The numbers behind the posture

3/50

Piston rings that passed Toyota’s first inspection, late 1930s

Employees, himself included, when Honda Motor was incorporated in 1948

30,000

Super Cubs a month: the sales target Fujisawa set in 1957

Top 5

Places taken in both the 125cc and 250cc Isle of Man TT races, 1961

~230

Patent applications pending on the CVCC engine by October 1972

62,000

Visitors to the thank-you gatherings held in place of a funeral, 1991

Tensions

Where the model strains

Failure is a lucky break, not a wasted effort

By his own account he got “really nasty” with staff whose work fell short

Racing exists to make cars safer for customers

His air-cooled RA302 crashed in its only race in July 1968, killing Jo Schlesser

Judge every design by what is new in it

Held to air cooling until his engineers and Fujisawa overruled him in 1969

President, with his name on the company

Fujisawa held the company seal, ran management and judged when both should retire

For operators

What to take from it

Find the partner who has what you lack, then stay out of their half.

Set the target at world level before you can afford it.

Treat a flop as proof of where not to go, and tell your team so.

When your own conviction blocks the company, choose the company.

Sources

Honda Motor, official history episodes at global.honda, in English translation (chapters for 1936, 1946, 1947, 1948, 1949, 1952, 1953, 1954, 1959, 1962, 1964, 1968, 1969, 1972 and 1991) · Honda Monthly (March 1953) and company newsletters (1959 to 1969), as quoted there · Honda Philosophy (global.honda) · Encyclopaedia Britannica · Wikipedia entries on Soichiro Honda, Takeo Fujisawa (English and Japanese), the Super Cub, the Cub F, the Juno, the RA302 and the 1959 and 1961 Isle of Man TT, citing Sakiya (1982) and Alexander (2008). All quotes are translations from Japanese. Quotes shortened.

All profilesJ. Robert Oppenheimer

Scientific director of Los Alamos · 1943 to 1945

J. Robert
Oppenheimer

The Open-Lab Director

He held that a hard problem is solved only where people are free to say what they know and do not know. He ran Los Alamos that way, then argued for the same openness between nations, and lost his clearance in 1954.

Drawn from his 1945 farewell speech at Los Alamos, his 1954 letter and testimony to the Atomic Energy Commission, the 1965 Groueff and NBC interviews, his letters and lectures (1932 to 1953), and histories of the project.

The central idea

Open inside the fence

Groves ran the Manhattan Project on need to know. Oppenheimer insisted that inside Los Alamos the scientific staff hear the whole problem, in opening lectures and a weekly colloquium.

Swipe sideways to see the full diagram

Groves’s rule, across the projectNeed to knowWorkers told only what their own job requiredNo discussing the work beyond designated supervisorsNearly 130,000 people at the project’s peak

Oppenheimer’s rule, inside the fenceFree to discuss what they know and do not knowFive opening lectures for 30 to 50 arrivals, 5 to 14 April 1943A weekly colloquium for the scientific staffA civilian laboratory: the plan for Army commissions was never carried out

From the first lecture on 5 April 1943 to the Trinity test on 16 July 1945.27 months

Peak employment is for the whole Manhattan Project. Lecture dates and attendance are for Robert Serber’s talks of April 1943, later issued as The Los Alamos Primer. The colloquium was limited to staff scientists.

Operating system

Eight principles

One place where people can talk

He argued for a single laboratory before he argued for leading it.

Oct 1942Groves appointed him on 19 October to lead one central laboratory for bomb design.

Nov 1942Toured sites with Groves on 16 November and steered the choice to the Los Alamos mesa.

“This thing will never get on the rails unless there is a place where people can talk to each other.”Stephane Groueff interview, 1965

Colloquium over compartments

Staff scientists heard the whole problem every week, against the project’s need-to-know rule.

Apr 1943Opened the lab with five lectures by Robert Serber that told 30 to 50 arrivals the full problem.

Jul 1944Told the staff on 4 July that reactor-made plutonium would not work in a gun-type bomb.

“Secrecy strikes at the very root of what science is, and what it is for.”Speech to Los Alamos scientists, 2 Nov 1945

Recruit on the mission

He could not offer rank or comfort, so he sold urgency, feasibility and a place in history.

1943Traveled the country to recruit; Bethe, Bacher, Teller and Feynman were among those who came.

1943Rabi and Bacher opposed Army commissions; the lab stayed civilian under the University of California.

“I had to rely on their sense of the interest, urgency, and feasibility of the Los Alamos mission.”Letter to the AEC, 4 March 1954

Direct from inside the work

Colleagues recalled a director who knew each technical problem and could summarise a room into agreement.

1943–45Victor Weisskopf wrote that he “did not direct from the head office.”

1982Aide David Hawkins recalled that after his summaries people found they agreed with each other.

“There was no electronic equipment to do that. We had to invent it.”Stephane Groueff interview, 1965

Reorganise when physics says so

When reactor plutonium ruled out the gun design, he rebuilt the laboratory around implosion within weeks.

Jul 1944Abandoned the Thin Man plutonium gun on 17 July and gave implosion overriding priority.

Aug 1944Created G Division under Robert Bacher and X Division under George Kistiakowsky.

“Time and again the laboratory drew itself together and faced the new problems and got on with the work.”Letter to the AEC, 4 March 1954

Speed as a duty

The job counted only if it was finished in time to affect the war.

1943–44The site grew from his early estimate of 100 staff to about 5,700 residents by the end of 1944.

Jul 1945Trinity fired on 16 July, 27 months after the first lectures; he had bet on 0.3 kilotons.

“We worked by night and by day; and in the end the many jobs were done.”Letter to the AEC, 4 March 1954

Do the work, then answer for it

He held that a scientist cannot stop such work, and cannot pretend afterwards that nothing changed.

Jun 1945Signed the Scientific Panel report that saw “no acceptable alternative to direct military use.”

Apr 1954Told the hearing that when something is “technically sweet, you go ahead and do it.”

“We knew the world would not be the same. A few people laughed, a few people cried.”NBC, The Decision to Drop the Bomb, 1965

Argue for openness in public

After 1945 he pressed for international control of atomic energy and for less secrecy at home.

Mar 1946Wrote much of the Acheson-Lilienthal Report, which proposed an international atomic authority.

Oct 1949Chaired the AEC advisory committee that unanimously opposed a crash hydrogen bomb program.

“We know that the wages of secrecy are corruption.”Science Talent Institute address, 1950

Evolution

How the thinking changed

YearsEraWhat they believedWhat changed it

1904–1929

Prodigy and student

Physics came first: “I need physics more than friends,” he told his brother.

Unhappy in the Cambridge laboratory, he moved to Göttingen and took a theory PhD under Max Born in 1927.

1929–1942

Berkeley teacher

Discipline needs a real end; he built a school of theoretical physics around students.

Ernest Lawrence drew him into bomb work in October 1941; by May 1942 he led the fast-neutron calculations.

1942–1945

Los Alamos

A bomb needs one laboratory where scientists talk freely and move fast.

Trinity worked on 16 July 1945. He later recalled the line “Now I am become Death,” and resigned on 16 October.

1945–1949

Adviser on control

Atomic weapons need international control; “the physicists have known sin.”

The Soviet Union rejected the Baruch Plan at the United Nations, then tested its own bomb in August 1949.

1949–1954

H-bomb and hearing

A crash hydrogen bomb program might weaken the United States rather than strengthen it.

Truman ordered the H-bomb in January 1950. The AEC revoked his clearance 4 to 1 on 29 June 1954.

1954–1967

Princeton years

Open access to knowledge keeps a technical world “a world of human community.”

Led the Institute for Advanced Study until 1966, took the Fermi Award in 1963, died on 18 February 1967.

Scale

The numbers behind the posture

Months from the first Los Alamos lectures, April 1943, to Trinity

100

His first staff estimate, about 1942: 50 scientists, 50 technicians

~4,000

Laboratory members by spring 1945, by his own 1954 account

0.3 kt

His pool bet on Trinity’s yield, July 1945; it measured about 18.6 kt

4 to 1

AEC vote against restoring his clearance, 29 June 1954

2022

Year the Department of Energy vacated the 1954 decision, 16 December

Tensions

Where the model strains

Open discussion inside the fence

Klaus Fuchs, in the Theoretical Division from August 1944, passed implosion data to Moscow

Asked officials and the public for candor

Admitted in 1954 that he had invented his 1943 account of the Chevalier approach

Signed the June 1945 advice for “direct military use”

Told Truman that October: “I feel I have blood on my hands”

Ruled a security risk in 1954, his clearance revoked

Vacated in 2022 for a flawed process, with no ruling on whether he should have had access

For operators

What to take from it

Put the whole problem in front of everyone who has to solve it.

Recruit on mission and urgency when you cannot offer rank or comfort.

When the facts kill the plan, reorganise in weeks and say so to the whole staff.

Tell your overseers the truth early. Late candor reads as concealment.

Sources

Oppenheimer, speech to the Association of Los Alamos Scientists (2 November 1945) · Letter to K. D. Nichols and testimony, In the Matter of J. Robert Oppenheimer (AEC, 1954) · Stephane Groueff interview, Voices of the Manhattan Project (1965) · NBC, The Decision to Drop the Bomb (1965) · Letter to Frank Oppenheimer (1932), in Letters and Recollections · MIT Arthur D. Little Lecture (1947) · “Encouragement of Science” (1950) · Science and the Common Understanding (1953) · Scientific Panel report (June 1945) · Atomic Heritage Foundation · David Hawkins interview (1982) · Bird and Sherwin, American Prometheus (2005), and Rhodes, The Making of the Atomic Bomb (1986), as cited in reference histories · Lyndon Johnson, Fermi Award remarks (1963) · Department of Energy statement (December 2022). Quotes shortened.

All profilesJamie Dimon

Chairman and CEO of JPMorgan Chase · 2006 to 2026

Jamie
Dimon

The Fortress Builder

He runs the bank to survive its worst year, not to win its best one. Capital that looks excessive in a boom made JPMorgan the buyer of failing rivals in 2008 and 2023. The rest is detail, candor and cost discipline.

Drawn from his letters to shareholders (2005 to 2025), the Acquired live interview (2025), the Q3 2025 earnings call, talks at the U.S. Chamber of Commerce and Norges Bank Investment Management (2026), and JPMorgan’s 2023 First Republic statement.

The central idea

The fortress balance sheet

Hold more capital and liquidity than a good year needs. It costs some return in a boom and makes JPMorgan the buyer when a rival fails.

Swipe sideways to see the full diagram

Step oneHold more than you needTangible common equity of $49B in 2005 and $280B in 2025. Liquid assets of $106B, then $1,464B.

→

Step twoAccept a duller boom“You might do worse than others in the good times, but you’re still there when the dust clears.”

→

Step threeBe the buyer2008: Bear Stearns for $1.5B and Washington Mutual for $1.9B. 2023: First Republic, with $10.6B paid to the FDIC.

→

Step fourAbsorb and compoundWashington Mutual brought 2,200 branches and $160B of deposits. First Republic brought $92B of deposits.

Larger earnings rebuild the capital, which feeds back to step one.

The result: compounded annual gain in JPMorgan’s stock from the Bank One merger on 1 July 2004 to the end of 2025, against 10.8% for the S&P 500.13.4% a year

Balance sheet figures are from the fortress balance sheet chart in the 2025 shareholder letter (April 2026). Bear Stearns and Washington Mutual figures are from the 2008 letter; First Republic figures are from JPMorgan’s statement of 1 May 2023. The quote is from the Acquired live interview, 2025. Stock returns are from the 2025 letter.

Operating system

Eight principles

A fortress, not a philosophy

Capital, liquidity and loss reserves are held above what a good year or a regulator requires.

2008Stayed profitable in the crisis year, earning nearly $6B with $24B of credit loss reserves.

2025Tangible common equity reached $280B, up from $49B in 2005.

“Having a fortress balance sheet is a strategic imperative, not a philosophical bent.”2007 shareholder letter

Buy when others must sell

Strength held in reserve is spent when a rival fails and regulators need a buyer.

2008Bought Bear Stearns for $1.5B in March and Washington Mutual’s bank for $1.9B in September.

May 2023Took on First Republic’s $173B of loans and $92B of deposits, paying the FDIC $10.6B.

“Our financial strength, capabilities and business model allowed us to develop a bid.”First Republic statement, May 2023 (shortened)

Bad things happen

He does not forecast the cycle. He prepares for a wide range of outcomes and warns while markets are calm.

Apr 2025Wrote that new U.S. tariffs would likely raise inflation and the odds of recession.

Oct 2025Wrote off $170M on Tricolor and warned: “When you see one cockroach, there are probably more.”

“We do not know exactly what will occur or when, but we do know that bad things happen.”2007 shareholder letter

Name the mistake yourself

Losses are disclosed and owned in public, including his own first dismissal of the problem.

2012Called early London Whale reports a “tempest in a teapot” in April; the loss reached $6.2B.

2013His 2012 pay was halved to $11.5M, and the bank paid $920M in fines that September.

“The London Whale was the stupidest and most embarrassing situation I have ever been a part of.”2012 shareholder letter

Kill bureaucracy, relentlessly

Size breeds complacency, politics and meetings. He treats removing them as permanent work.

Apr 2025His letter named complacency, arrogance, bureaucracy and “BS” as what kills companies.

Apr 2026Told an Oslo audience he cancels meetings when information has not been shared properly.

“You must kill bureaucracy all the time and relentlessly.”2024 shareholder letter

Nothing is too small

He stays in the facts and the numbers, because at the top bad news arrives filtered.

Nov 2025Said he still reads customer complaints himself because his staff filters too much.

Jul 2026Said the bigger the job, the less you know, and that slides made to look good hide bad news.

“Details, facts, analysis, no bullshitting, no meetings after meetings.”U.S. Chamber of Commerce, January 2026

Work is learned in person

Apprenticeship and attention need people together. He set the office rule and did not negotiate it.

2025Ordered staff back five days a week; more than 1,200 signed a petition against it.

Oct 2025Opened a $3B, 60-story headquarters at 270 Park Avenue for 10,000 employees.

“You can’t learn working from your basement.”Bloomberg interview, 2025, via Fortune

Cut waste, keep investing

Costs that do not serve clients are cut so that spending on branches, people and systems continues.

2005–06Unallocated corporate overhead fell from $2.4B to $750M after the Bank One merger.

2021–25Opened more than 900 branches and added more than 60,000 employees in five years.

“It’s easy to grow short-term earnings: just stop investing in your company’s future.”2006 shareholder letter (shortened)

Evolution

How the thinking changed

YearsEraWhat they believedWhat changed it

1982–1998

Weill’s apprentice

Buy unloved financial firms, cut costs and merge upward.

Followed Sandy Weill from American Express to Commercial Credit, Travelers and Citigroup. Weill fired him in November 1998.

2000–2004

Bank One turnaround

A CEO should go up or down with the ship he runs.

Put about $60M, half his net worth, into Bank One shares in March 2000. Merged the bank into JPMorgan Chase in July 2004.

2004–2007

Building the fortress

Costs are too high and returns too low; fix both before the cycle turns.

CEO from the end of 2005. Cut overhead and declined option-ARM mortgages, yet subprime and home equity losses beat his estimates.

2008–2011

Crisis buyer

A strong balance sheet earns the right to buy failing rivals.

Bear Stearns and Washington Mutual in 2008. He later tied about 70% of nearly $19B in mortgage legal costs to the two deals.

2012–2019

Chastened operator

Controls must match risk, and problems do not age well.

The $6.2B London Whale loss and $920M in fines. He apologized to shareholders, and was treated for throat cancer in 2014.

2020–2026

Statesman, succession

Prepare for a wide range of outcomes; war, debt and cyber risk top the list.

Heart surgery in 2020, First Republic in 2023, $57.0B profit in 2025. Co-presidents named June 2026; he plans a few more years.

Scale

The numbers behind the posture

$57.0B

Net income in 2025, on revenue of $185.6B

13.4%

Annual stock gain, July 2004 to end 2025; the S&P 500 returned 10.8%

~$970B

Market value on 17 August 2026, up from $138B at the end of 2005

$280B

Tangible common equity in 2025, up from $49B in 2005

$6.2B

London Whale trading loss, 2012

$43M

His pay for 2025, up 10.3% on 2024

Tensions

Where the model strains

Be the buyer when a rival fails

2014 letter: “we would not do something like Bear Stearns again”

A fortress of capital and risk controls

The London Whale lost $6.2B in 2012 after he called it a “tempest in a teapot”

Says he has a few years left as CEO, then perhaps executive chairman

20 years as CEO; Marianne Lake, long a leading contender, announced her exit in June 2026

“Do the right thing, not necessarily the easy or expedient thing”

JPMorgan paid $290M and $75M in 2023 to settle claims over banking Jeffrey Epstein

For operators

What to take from it

Hold more capital and cash than a good year needs. It buys you options in a bad one.

Price the liabilities you inherit in a rescue deal, not only the assets.

Disclose your own mistake first, with the number attached.

Treat bureaucracy as a cost that regrows. Cut it every year.

Sources

JPMorgan Chase letters to shareholders (2005, 2006, 2007, 2008, 2012, 2014, 2023, 2024, 2025) · JPMorgan statement on First Republic (May 2023) · JPMorgan leadership pages (2026) · Acquired live interview via Fortune (2025) · Q3 2025 earnings call via Fortune · U.S. Chamber of Commerce (January 2026) · Norges Bank Investment Management conference (April 2026) · Master Investor Podcast via Fortune (July 2026) · Fortune (2025, 2026) · Banking Dive (2026) · Wikipedia for dates and the 2012 trading loss. Quotes shortened.

All profilesShigeru Miyamoto

Creator of Mario and Zelda at Nintendo · 1977 to 2026

Shigeru
Miyamoto

The Feel-First Designer

He starts with how a game feels in the hand, not with a story or a market. He takes ideas from ordinary life, watches ordinary people play, and overturns the work late if the fun is missing.

Drawn from Nintendo’s Iwata Asks interviews (2006 to 2012), David Sheff’s Game Over (1993), Nintendo’s 2023 Zelda film announcement and its 2026 governance report, sales data and first-quarter results.

The central method

Start from the feel

He gets the movement feeling good before there is a game around it. Story and setting are fitted afterwards, people are watched as they play, and weak work is overturned.

Swipe sideways to see the full diagram

Step oneNotice a feelingA cave he explored alone as a boy near Sonobe is the feeling he wanted players to have in The Legend of Zelda (1986).

→

Step twoMake it feel good to moveThe movement comes before the game. “Mario is a fun game, even if you just move him around,” he said in 2007.

→

Step threeWatch people playFor Super Mario Galaxy (2007), children, parents and people over 50 played in the office while staff observed their faces.

→

Step fourUpend the tea tableIf the fun is missing, change it, even late. The opening village of Twilight Princess (2006) was reworked after E3.

A late change sends the team back to step two.

The same method at one company: Donkey Kong (1981), Super Mario Bros. (1985), The Legend of Zelda (1986), Super Mario 64 (1996), Super Mario Galaxy (2007).1977 to 2026

Examples are from David Sheff’s Game Over (1993) and Nintendo’s Iwata Asks interviews on Twilight Princess (2006) and Super Mario Galaxy (2007). The quote is Nintendo’s English translation from Japanese. Game dates are first release.

Operating system

Eight principles

Feel before everything

Get the character’s movement right first, then build the game around what feels good.

1985Tested a Mario twice the normal size before designing levels. It felt good, so it stayed.

1996Super Mario 64 let players move Mario freely in 3D, which he called the good part of the game.

“As it felt really good, we continued to develop the idea.”Iwata Asks: New Super Mario Bros. Wii, 2009

Form follows function

Characters and objects are shaped by what the hardware allows and what the player must read.

1981Drew Mario with a moustache for Donkey Kong because it saved pixels and read clearly.

2007Emailed the whole Super Mario Galaxy team that an object’s function must be easy to understand.

“It’s very important to design these things so its function would be easily understood.”Iwata Asks: Super Mario Galaxy, 2007

Exaggerate ordinary life

His worlds come from walks, caves and things he passed, enlarged until they feel like adventure.

1985A manhole cover he passed as a student in Kanazawa became a way into hidden rooms in Super Mario Bros.

1986Zelda’s caves and mazes drew on a cave near Sonobe and the sliding doors of his family home.

“I exaggerate what I experience and what I see.”David Sheff, Game Over, 1993

One idea, several problems

A good idea removes several obstacles at once. He starts by listing what cannot be done.

1981Gave Mario a jump in Donkey Kong; without it, he said, the game would have been far too hard.

1983Higher jumps in Mario Bros. needed a new enemy: a turtle that flips when hit from below.

“An idea is something which solves multiple issues at once.”His definition, Iwata Asks: Mario 25th, 2010

Watch people play

Written feedback can mislead. He watches faces, and tests on people who do not play games.

1985Placed the first enemy in Super Mario Bros. so that new players learn to jump unprompted.

2007Had children, parents and people over 50 play Super Mario Galaxy in the office, and observed.

“He wrote that the game is fun, but he didn’t look like he was having fun.”Iwata Asks: Super Mario Galaxy, 2007

Upend the tea table

When work falls short he overturns it, even late. Staff call it upending the tea table.

2006Had the opening village of Twilight Princess reworked after its E3 showing.

2015–16Star Fox Zero, which he produced, slipped from November 2015 to April 2016.

“We can go ahead and change this, don’t you think?”Iwata Asks: Mario 25th Anniversary, 2010

Play before story

The experience of playing is the story. Plot is fitted to the system, or left out.

1998Built Ocarina of Time’s system first and set the story to what the system could hold.

2012Told the Paper Mario: Sticker Star team to cut the story back and use only Mario characters.

“It’s fine without a story, so do we really need one?”Quoted by staff, Iwata Asks: Sticker Star, 2012

Beyond the console

Since 2015 he has moved Nintendo’s characters into theme parks, a museum and films.

Mar 2021Super Nintendo World, which he helped design, opened at Universal Studios Japan.

Apr 2026The Super Mario Galaxy Movie, his second as producer, opened and passed $1B by July.

“I have been working on the live-action film of The Legend of Zelda for many years now.”Nintendo announcement, November 2023

Evolution

How the thinking changed

YearsEraWhat they believedWhat changed it

1952–1977

Sonobe and Kanazawa

Exploring is its own reward. He wanted to be a puppeteer or a painter, and drew cartoons.

Studied industrial design from 1970. His father arranged a meeting with Hiroshi Yamauchi, who hired him in 1977.

1977–1984

Artist to designer

Design is form fitted to function, and to a handful of pixels.

Told in 1980 to rescue the failed arcade game Radar Scope, he made Donkey Kong (1981). In 1984 he was given his own team.

1984–1995

Mario and Zelda

Worlds should hide rewards for players curious enough to look.

Super Mario Bros. (1985), Zelda (1986) and Super Mario World (1990) made him the best-known designer in games.

1996–2005

3D and producing

Moving freely in 3D is fun before there is a game.

Super Mario 64 (1996) and Ocarina of Time (1998). He moved to producing and became a Representative Director in 2002.

2005–2015

Games for everyone

Parents, grandparents and non-players should be able to play together.

Wii (2006) sold 101.63M consoles; its successor Wii U sold 13.56M. Satoru Iwata died in July 2015.

2015–2026

Fellow and producer

Nintendo’s characters can live in films, parks and a museum as well as games.

Parks from 2021, a museum in 2024, Switch 2 in 2025 and two Mario films past $1B each. A Zelda film is due in 2027.

Scale

The numbers behind the posture

Years at Nintendo, 1977 to 2026

800M+

Mario games sold, the best-selling game franchise (Wikipedia, 2026)

$1.36B

The Super Mario Bros. Movie, worldwide gross, 2023

$1.01B

The Super Mario Galaxy Movie, worldwide gross to July 2026

101.6M

Wii consoles sold, life to date (Nintendo, June 2026)

¥246M

His pay as a Nintendo director for the year to March 2026

Tensions

Where the model strains

Says “genius” is the wrong word and that he sees “as a normal person”

At 73 he is still a Representative Director, paid ¥246M in the year to March 2026

Told a team “It’s fine without a story”

Reviewers faulted Paper Mario: Sticker Star (2012) for its thin story and generic cast

Upends the tea table to protect quality

Star Fox Zero, delayed five months, scored 69 on Metacritic; its controls were faulted

Said in 2010 he lacked “what it takes” to work at Hollywood scale

Two Mario films grossed about $2.4B; the 2026 one scored 42% on Rotten Tomatoes

For operators

What to take from it

Prototype the feel first. Add features only to what already feels good.

Watch first-time users’ faces, not just their written feedback.

Look for the one idea that removes several problems at once.

Overturn weak work late if you must, and budget for the delay.

Sources

Nintendo, Iwata Asks: Twilight Princess (2006), Super Mario Galaxy (2007), New Super Mario Bros. Wii (2009), Super Mario Bros. 25th Anniversary (2010), Ocarina of Time 3D (2011), Paper Mario: Sticker Star (2012) · David Sheff, Game Over (1993) · Nintendo, Zelda film announcement (November 2023) · Nintendo corporate governance report (June 2026) · Nintendo sales data and first-quarter results (August 2026) · Box office, review scores and dates via Wikipedia (October 2026). Quotes are English translations from Japanese. Quotes shortened.

All profilesLarry Ellison

Co-founder, chairman and CTO of Oracle · 1977 to 2026

Larry
Ellison

The Contrarian Closer

He looks for the place where expert opinion is wrong, commits before the incumbent does, and outsells everyone. He did it to IBM with the relational database from 1977 and is doing it again, on borrowed money, with AI data centres.

Drawn from his Smithsonian oral history (1995), his Academy of Achievement interview (1997), his remarks on cloud computing (2008), Oracle’s earnings releases (2025 to 2026), and reporting by Forbes, The Register and TechCrunch.

The central play

Ship the incumbent’s idea first

IBM invented the relational database and published the research. Ellison read it, shipped a product before IBM did, outsold everyone, and later bought the rivals that were left.

Swipe sideways to see the full diagram

1 · ReadThe incumbent’s researchTed Codd of IBM published the relational model in 1970. Ellison read IBM’s follow-up research in November 1976.

→

2 · DoubtThe expert viewConventional wisdom held that relational databases would be too slow to sell.

→

3 · ShipBefore the inventorOracle Version 2 went on sale in 1979. IBM announced its own relational product, SQL/DS, in 1981.

→

4 · SellTake the shareSales roughly doubled every year for ten years. By 1995 he claimed more share than all rivals combined.

→

5 · BuyThen the rivalsPeopleSoft for $10.3B in 2005, Sun for about $7.4B in 2010, Cerner for $28.3B in 2022.

He is running steps one to three again on AI data centres. Oracle’s contracted backlog went from $455B in August 2025 to $664B in August 2026, ahead of the cash to build them.

Steps one to four follow his own account in his 1995 Smithsonian oral history and 1997 Academy of Achievement interview. Deal values are as announced, with the year each closed. Backlog is remaining performance obligations in Oracle’s results for the quarters ended 31 August 2025 and 31 August 2026.

Operating system

Eight principles

Find where the experts are wrong

An edge exists only where accepted opinion is mistaken, so he goes looking for the mistake.

1976Read IBM’s relational research when the accepted view was that it would be too slow to sell.

1995Bet against the PC with a network computer priced under $500. It never sold in volume.

“Whenever you can find flaws in conventional wisdom is when you get real competitive advantages.”Smithsonian oral history, 1995

Beat the inventor to market

He treated IBM’s published papers as a specification and raced IBM to ship the product.

1977Started the company with $2,000, of which $1,200 was his.

1979Shipped Oracle Version 2; there was no Version 1. First customer, by his account: Wright-Patterson Air Force Base.

“Oracle decided to see if we couldn’t beat IBM to market with this technology.”Academy of Achievement, 1997

Market share is everything

Take the market first. In his view profit comes from share, not the other way round.

1980sSales roughly doubled every year for ten years, by his account.

1995Said Oracle had more database share than all of its competitors combined.

“Market share is everything.”Smithsonian oral history, 1995

Change the managers, keep the bet

When growth broke the company in 1990 he replaced the people around him, not the strategy.

1990First quarterly loss. About 400 jobs went, 10% of staff, after sales were booked too early.

1991–92Brought in Jeff Henley as finance chief and Ray Lane to run operations.

“The company had outgrown the management.”Academy of Achievement, 1997

Buy the rivals

Once enterprise software matured he stopped out-building competitors and bought their customers.

2003–04Pursued PeopleSoft with a hostile bid for 18 months and won it for $10.3B.

2010–22Bought Sun for about $7.4B, then Cerner for $28.3B, Oracle’s largest deal.

“It is my job to go out into the marketplace and win.”Academy of Achievement, 1997

Mock the fashion, then build it

He ridiculed the cloud label in 2008, then spent the next decade rebuilding Oracle around it.

2008Told analysts that cloud computing was a fashion, not a new technology.

FY2026Cloud revenue reached $34.0B, about half of Oracle’s $67.4B total.

“It’s complete gibberish. It’s insane. When is this idiocy going to stop?”On cloud computing, Oracle OpenWorld, 2008

Keep the engineer’s job

He gave away the CEO title twice over but never the technology role or the chair.

Sep 2014Stepped down as CEO after 37 years and stayed on as chairman and chief technology officer.

Sep 2025Named Clay Magouyrk and Mike Sicilia co-CEOs. Safra Catz became executive vice chair.

“I was an engineer and a builder by nature.”Smithsonian oral history, 1995

Nothing riskier than no risk

At 80 he committed Oracle to build AI data centres faster than its cash flow could fund them.

Jan 2025Announced Stargate at the White House with OpenAI and SoftBank, a $500B plan.

FY2026Capital spending rose to $55.7B from $21.2B. Free cash flow was minus $23.7B.

“There is really nothing riskier than not taking risks.”Smithsonian oral history, 1995

Evolution

How the thinking changed

YearsEraWhat they believedWhat changed it

1944–1977

Chicago to Ampex

Authority and experts deserve to be doubted.

Left two universities without a degree, worked as a programmer in California, and read IBM’s relational research in 1976.

1977–1990

Land grab

Ship first, sell everywhere, and share will pay for it.

Oracle went public in March 1986. Then sales booked too early led to its first loss.

1990–2002

Rebuild, then network

The network, not the PC, is the computer.

New managers restored controls. The network computer failed to sell, though the internet lifted database demand.

2003–2014

Consolidator

Enterprise software will consolidate, so be the buyer.

PeopleSoft, Siebel, BEA and Sun. He dismissed cloud in 2008 and gave up the CEO title in September 2014.

2014–2024

Chairman and CTO

Oracle can still win in cloud, late, with a better design.

Bought NetSuite and Cerner, put the database inside Microsoft, Google and Amazon clouds, and moved the head office to Texas.

2025–2026

The AI build-out

AI demand is bigger than Oracle’s cash flow, so borrow and build.

Backlog hit $664B and he was briefly the richest person. Then the shares fell, S&P cut the rating and 21,000 jobs went.

Scale

The numbers behind the posture

$664B

Contracted backlog (RPO), August 2026, up from $455B a year earlier

$55.7B

Capital spending in fiscal 2026, up from $21.2B in fiscal 2025

$393B

Net worth on 10 September 2025, briefly the world’s richest person

$176B

Net worth on 29 September 2026, ranked eighth (Forbes)

~$40B

His personal guarantee for Paramount’s Warner Bros. bid, December 2025

15%

Oracle’s stake in TikTok’s US joint venture, formed January 2026

Tensions

Where the model strains

Called cloud computing “complete gibberish” in 2008

Cloud was $34.0B of Oracle’s $67.4B revenue in fiscal 2026

“We never, ever lost money,” he said of Oracle’s first years

Free cash flow of minus $23.7B in fiscal 2026, and an S&P rating of BBB−

A record $664B of signed contracts

About $300B of it is one customer, OpenAI, which is not yet profitable

Credits his success to questioning authority

Said in 2024 that AI-monitored cameras would keep citizens in line

For operators

What to take from it

Read the incumbent’s published research, then ship it before they do.

Take share first, but never book a sale before it is real.

If you mocked a shift that proved real, reverse and commit fully.

Count how much of your backlog depends on a single customer.

Sources

Smithsonian Institution oral history (1995) · Academy of Achievement interview (1997) · Remarks at Oracle OpenWorld (2008) · Oracle earnings releases (September 2025 to September 2026) · Forbes (2026) · The Register (2024 to 2026) · TechCrunch (2025, 2026) · Wikipedia entries on Oracle, Stargate, Paramount Skydance and TikTok’s US joint venture. Quotes shortened.

All profilesMichael Dell

Founder, chairman and CEO of Dell · 1984 to 2026

Michael
Dell

The Direct Operator

He cut out the dealer, built each machine after it was ordered and collected the cash before paying suppliers. When the model aged he rebuilt the company around it, once from the CEO’s chair and once in private.

Drawn from his Academy of Achievement interview (2008), Forbes on the 2013 buyout, McKinsey Quarterly (2024), David Senra (2025), TIME (2026), his memoirs Direct from Dell (1999) and Play Nice But Win (2021), and Dell Technologies’ results to September 2026.

The central loop

The direct model

Take the order first, build the machine second, pay suppliers last. He called the result a negative cash conversion cycle: customers’ money, not investors’, funded the growth.

Swipe sideways to see the full diagram

Step oneOrder directNo dealer in between. By phone from 1984, on dell.com from 1996.

→

Step twoBuild to orderRivals’ supply chains held about 90 days of stock. Dell got to about five, by his account.

→

Step threeShip and collectThe customer pays at shipment, before Dell has paid for the parts.

→

Step fourPay suppliers later“We actually collect money way before we pay the money out.”

→

Step fiveGrow on the floatCustomers’ cash funds expansion: $1,000 of capital in 1984, a $30M IPO in 1988.

Direct orders also show what buyers want next, which feeds back to step one.

The same order-first pattern in 2026: AI servers booked by customers and not yet shipped.$95B backlog

Sequence and inventory days as Michael Dell described them to the Academy of Achievement (2008) and David Senra (2025). Capital and IPO figures are for 1984 and June 1988. AI server backlog as reported by Dell Technologies on 1 September 2026.

Operating system

Eight principles

Sell direct, build to order

Remove the dealer, take the order first, then build exactly what was asked for.

1984Started PC’s Limited from a University of Texas dorm room with $1,000.

Mar 1997Dell.com, opened in 1996, was selling about $1 million of computers a day.

“What if you could sell directly to the end customer and do it way more efficiently?”Academy of Achievement, 2008

Collect before you pay

Customers pay at shipment and suppliers are paid later, so growth largely funds itself.

Jun 1988Raised about $30 million in the IPO at $8.50 a share, four years after starting with $1,000.

2002Operating costs were 10% of $35 billion in revenue, against 21% at Hewlett-Packard.

“We had what’s known as a negative cash conversion cycle, which is a very good thing.”Academy of Achievement, 2008

Curiosity, with room for error

He learns by opening things up and trying them, and accepts that some attempts fail.

1980At 15, bought an Apple II and took it apart to see how it was built.

Sep 2024Called Dell’s attempt at mobile phones a turn it missed and “a bad use of capital.”

“I think you have to be able to experiment and make mistakes.”Academy of Achievement, 2008

Founders can change everything

He believes a founder has a licence to remake a company that a hired manager lacks.

Jan 2007Returned as CEO when Kevin Rollins resigned after four quarters of missed earnings.

2009–13Bought more than 20 companies for about $13 billion, including Perot Systems for $3.9 billion.

“A founder has special permission to make sweeping changes across an organization.”Forbes, 2013

Go private to transform

Away from quarterly scrutiny, and with cheap debt, he moved Dell from PCs toward enterprise systems.

Oct 2013Closed a buyout of about $25 billion with Silver Lake and came out owning roughly 75%.

Sep 2016Still private, closed the $67 billion EMC deal, funded with $45.9 billion of new debt.

“When money goes on sale, stuff happens.”McKinsey Quarterly, 2024

Play nice but win

The title of his 2021 memoir: fight hard for the outcome and keep it about the business.

2013Carl Icahn took a $1 billion stake to block the buyout; Dell raised the price and won the vote.

2018Icahn sued over the return to public markets; Dell raised its offer to tracking-stock holders.

“It’s a big poker game to him. It’s not about the customers. It’s not about the people.”On Carl Icahn, Forbes, 2013

Build for the next workload

Each platform shift, from PCs to servers to AI, is treated as hardware to be built to order.

Feb 2026Closed fiscal 2026 with $64 billion of AI server orders and revenue of $113.5 billion.

Sep 2026Reported $60.9 billion of AI server orders in one quarter and a $95 billion backlog.

“In the tech industry, we often say ‘change or die.’ There’s the quick or the dead.”McKinsey Quarterly, 2024

Give capital, not advice

His largest gift puts a small investment stake in children’s names and leaves the choices to families.

Dec 2025Pledged $6.25 billion with Susan Dell: $250 each for 25 million children’s “Trump accounts.”

Apr 2026Took the couple’s lifetime giving to the University of Texas at Austin past $1 billion.

“We have too many families that have no capital investment at all.”TIME, 2026

Evolution

How the thinking changed

YearsEraWhat they believedWhat changed it

1984–1988

Dorm room to IPO

A PC sold direct can be better and cheaper than one sold through a dealer.

Sales outgrew his $1,000 start. He left the University of Texas at 19 and took the company public in June 1988.

1989–1995

Growing pains

Growth solves most problems.

Excess inventory and the cancelled Olympic line in 1989, then a loss in 1993, put liquidity and profit ahead of growth.

1996–2004

The direct machine

Direct selling plus the internet beats any channel.

Dell.com hit $1M a day in 1997 and Dell passed Compaq in 2001. He handed the CEO job to Kevin Rollins in 2004.

2004–2012

Stall and return

The model would keep winning on cost.

HP took the top PC spot in 2006 and earnings were restated. Back as CEO in 2007, he cut 10% of jobs; a $4M SEC penalty followed.

2013–2018

Private

A rebuild needs owners who can ignore the quarter.

A $24.4B buyout in 2013 and the $67B EMC deal in 2016, then a contested return to the NYSE in December 2018.

2019–2026

Public again, then AI

Infrastructure built to order wins each new computing wave.

VMware spun off in 2021. AI server orders pushed fiscal 2027 revenue guidance to $192B. He remains chairman and CEO.

Scale

The numbers behind the posture

$1,000

Starting capital in his University of Texas dorm room, 1984

$24.4B

Buyout with Silver Lake to take Dell private, announced February 2013

$67B

EMC acquisition, announced October 2015 and closed September 2016

$95B

AI server backlog reported on 1 September 2026

~97K

Employees in early 2026, down from about 133,000 in early 2023

$6.25B

Pledged with Susan Dell for 25 million children’s accounts, Dec 2025

Tensions

Where the model strains

Built the company on cutting out the middleman

By 2013 Dell had more than 140,000 channel partners producing about $16B a year

Left public markets in 2013, saying investors saw only “a PC company”

Returned in 2018 via a tracking-stock deal that Icahn sued over until the offer rose

“AI is going to augment humans, not replace them,” he said in 2024

Dell’s headcount fell from about 133,000 in early 2023 to about 97,000 in early 2026

Calls the $6.25B gift seed capital for families who have none

CNBC tied Dell’s $9.7B Pentagon contract in May 2026 to his closeness to Trump

For operators

What to take from it

Get paid before you pay. A negative cash cycle lets growth fund itself.

Sell to the buyer directly. The order book is your best market research.

If public markets block a rebuild, change the owners, not the plan.

Treat your model as a tool. Drop parts of it when customers move.

Sources

Academy of Achievement interview (2008) · Forbes (2013) · McKinsey Quarterly (2024) · David Senra podcast (2025) · TIME100 Philanthropy (2026) · Michael Dell, Direct from Dell (1999) and Play Nice But Win (2021) · Dell Technologies results (February, May and September 2026), press release (May 2026) and leadership page · The Franklin Institute (2014) · The Register (2025) · CNBC (May 2026) · Wikipedia entries on Michael Dell, Dell and Trump accounts, for dates. Quotes shortened.

All profilesAkio Morita

Co-founder of Sony · 1946 to 1994

Akio
Morita

The Market Maker

He held that customers cannot ask for what they have never seen. Sony built the product first, put its own name on it, and taught the public to want it, in Japan and then from inside its foreign markets.

Drawn from his memoir Made in Japan (1986), Masaru Ibuka’s founding prospectus (1946), his Playboy interview (1982), his essays in The Japan That Can Say No (1989), Time’s 1998 profile and John Nathan’s Sony: The Private Life (1999).

The central idea

Create the market

Sony did not ask the public what it wanted. It decided what was possible, built it under its own name, and then taught people why they needed it.

Swipe sideways to see the full diagram

“The public does not know what is possible, but we do.”

Step oneSkip the surveyRefine the idea of the product and its use inside the company. Do not ask the public what it wants.

→

Step twoBuild what nobody asked forA radio for the shirt pocket in 1957. A tape player that could not record in 1979.

→

Step threePut the Sony name on itIn 1955 he refused Bulova’s order for 100,000 radios because they would carry Bulova’s name.

→

Step fourTeach the publicWalkman, 1979: Sony forecast 5,000 a month and sold more than 30,000 in two months.

Quotation and method from Made in Japan (1986), where he wrote that Sony would “create a market” for a product “by educating and communicating” with the public. Bulova order, 1955. Walkman forecast and sales, July and August 1979.

Operating system

Eight principles

Lead the public

Customers cannot ask for what they have never seen, so decide what to build and then teach them its use.

1957Sold Sony’s small transistor radio as “pocketable”, his own word for a use nobody had requested.

1979Sony forecast 5,000 Walkmans a month and sold more than 30,000 in the first two months.

“The public does not know what is possible, but we do.”Made in Japan, 1986

The name before the order

A large order is not worth taking if it hides the maker’s name.

1955Refused Bulova’s order for 100,000 radios because they would carry the Bulova name.

1958Renamed the company Sony so that the company name and the brand name were the same word.

“Fifty years from now I promise you that our name will be just as famous as your company name is today.”To Bulova’s buyer, 1955; Made in Japan, 1986

Invention is not enough

A product people do not understand will not sell. Finding its users is the company’s job.

1950Japan’s first tape recorder sold slowly until he showed it to the Supreme Court, which bought twenty.

1960Founded Sony Corporation of America so that Sony ran its own sales in the United States.

“Unique products are not enough to keep a business going. You have to sell the products.”Made in Japan, 1986

Live inside the market

He would not sell to a country from a distance. He moved there, then built there.

1963Moved his wife and three children to New York to learn how Americans lived and bought.

Aug 1972Opened a plant in San Diego where American workers assembled Trinitron sets.

“I would have to move my family to the United States and experience the life of an American.”Made in Japan, 1986

People make the business

He treated the company as a family with a shared fate and judged people by work, not diplomas.

1966Published Never Mind School Records, arguing that school records should not decide a career.

1982Told Playboy that in a Japanese company workers and management “are in the same boat.”

“No theory or plan or government policy will make a business a success; that can only be done by people.”Made in Japan, 1986

Don’t repeat a mistake

Look for the cause of an error, not for someone to blame, and include your own errors.

1970sQuit calculators in a price war, then wrote that it “showed a lack of technical foresight on my part.”

1975–88Betamax lost to VHS. He wrote that Sony should have gathered a larger “family” of makers behind it.

“If you make a mistake, you will learn from it. Just don’t make the same mistake twice.”Made in Japan, 1986

An ideal factory for engineers

Ibuka’s charter put engineers’ freedom first. Morita found the money and sold what they made.

1953Committed $25,000 to a transistor licence that needed government approval to pay.

1968Launched Trinitron after selling only 13,000 sets of its failed predecessor, Chromatron.

“To establish an ideal factory that stresses a spirit of freedom and open-mindedness.”Founding prospectus, 1946 (Ibuka; Sony’s translation)

Ten years, not ten minutes

He told Americans they managed for the next quarter, and later told Japan its own model had to change.

1989Co-wrote The Japan That Can Say No with Shintaro Ishihara, faulting US firms for short horizons.

Feb 1992Wrote in Bungei Shunju that Japanese management was at a critical moment and people should work less.

“You seem to be concerned only with profits ten minutes from now.”The Japan That Can Say No, 1989 (unofficial translation)

Evolution

How the thinking changed

YearsEraWhat they believedWhat changed it

1921–1945

Heir and physicist

As first son he would become the fifteenth head of the family sake brewery.

Studied physics at Osaka, graduated in 1944, joined the navy and met Masaru Ibuka on a wartime research committee.

1946–1954

Tokyo Tsushin Kogyo

A unique product, well made, will find its buyers.

The 1950 tape recorder sold slowly until he showed it to Japan’s Supreme Court, which bought twenty machines.

1955–1962

The Sony name

A small firm must sell under its own name, even at the cost of orders.

Refused Bulova in 1955, sold the pocketable radio in 1957, became Sony in 1958 and set up Sony America in 1960.

1963–1975

The American years

To sell in America, live there and build there.

Moved to New York in 1963, launched Trinitron in 1968, became president in 1971, opened the San Diego plant in 1972.

1976–1988

Chairman

New products create their own markets, but formats need allies and content.

The Walkman beat its forecast, Betamax lost to VHS, and Sony agreed to buy CBS Records for $2B in 1987.

1989–1999

Hollywood and illness

Hardware needs software, and Japan should speak plainly to America.

Columbia Pictures in 1989, a stroke on 30 November 1993, resignation in 1994 and death on 3 October 1999.

Scale

The numbers behind the posture

15th

Generation of the family sake brewery he was born to inherit, 1921

¥190K

Founding capital of Tokyo Tsushin Kogyo, May 1946

100K

Radios Bulova ordered in 1955, refused because of the name

30,000

Walkmans sold in the first two months, 1979; forecast 5,000 a month

$2B

Price agreed for CBS Records, November 1987

$2.7B

Write-off on Sony’s film studios, November 1994

Tensions

Where the model strains

Lead the public with better technology

Betamax, launched in 1975, lost to VHS; Sony began selling VHS recorders in 1988

Plan ten years ahead, not ten minutes

Columbia, bought in 1989 on his “always dreamed of” wish, led to a $2.7B write-off in 1994

Sat on the Japan-U.S. Economic Relations Group, the “Wise Men”

He withdrew his essays from the 1991 English edition of The Japan That Can Say No

Judge people by ability, not school records

In 1982 Sony had no female executives; he told Playboy “Not yet” and that he wanted some

For operators

What to take from it

Don’t ask customers to imagine the product. Build it, then teach them its use.

Refuse the order that puts someone else’s name on your work.

Live in your largest market before you try to sell to it.

When a bet fails, find the cause, say it was yours, and don’t repeat it.

Sources

Akio Morita with Edwin Reingold and Mitsuko Shimomura, Made in Japan (1986) · Sony founding prospectus (1946), in Sony’s English translation · Playboy interview (August 1982) · The Japan That Can Say No (1989), unauthorised English translation · Kenichi Ohmae, Time (1998) · John Nathan, Sony: The Private Life (1999), via Business Week and Newsweek (October 1999) · Los Angeles Times (1987, 1994) · Britannica · Wikipedia. The prospectus and the 1989 essays are quoted in translation. Quotes shortened.

All profilesMasayoshi Son

Founder, chairman and CEO of SoftBank · 1981 to 2026

Masayoshi
Son

The 300-Year Gambler

About once a decade he picks the next computing platform and stakes SoftBank on it with borrowed money. Alibaba and Arm paid for the dot-com crash and WeWork. OpenAI and superintelligence are the largest bet yet.

Drawn from his messages in SoftBank Group’s reports (2021 to 2026), his shareholder meeting presentations (2024 to 2026), SoftBank earnings materials and releases (2019 to 2026), the DealBook conference (2020) and Lionel Barber’s Gambling Man (2024), most of it in SoftBank’s English versions of Japanese originals.

The central strategy

A new platform every decade

About once a decade he moves SoftBank onto the next computing platform and puts the balance sheet behind it. The sums have grown from $20M for Alibaba to $64.6B for OpenAI.

Swipe sideways to see the full diagram

1981PC softwareFounded SoftBank in September 1981 to distribute packaged software.

→

1995 to 2000InternetBacked Yahoo in 1995, then paid $20M for a stake in Alibaba in 2000.

→

2001 to 2013Broadband and mobileVodafone Japan for ¥1.75T in 2006. Control of Sprint in 2013, first priced at $20.1B.

→

2016 to 2017Chips and capitalArm for £24B in 2016. The $100B Vision Fund in 2017.

→

2025 to 2026ASIArtificial superintelligence. $64.6B into OpenAI for about 13%, with Stargate, Ampere and ABB Robotics.

What arriving early has cost: the fall in SoftBank’s market value after the internet peak of 2000, from about $180B to $2.5B.98% lost

Prices as announced: Alibaba (January 2000), Vodafone Japan (March 2006), Sprint (October 2012 offer, closed July 2013), Arm (2016), Vision Fund (2017). OpenAI total from SoftBank’s release of 27 February 2026, with the last tranche due 1 October 2026. Market value figures as reported by TechCrunch, November 2025.

Operating system

Eight principles

Move to the next platform

About once a decade he shifts SoftBank’s weight to the next computing platform, before the last one peaks.

2006Bought Vodafone Japan for ¥1.75T to move from broadband into mobile.

2025–26Committed $64.6B to OpenAI by October 2026, for a stake of about 13%.

“Its significance may even surpass the impact of the internet and smartphones.”Message from the Chairman & CEO, 2026

Plan in decades and centuries

He writes the long plan first: a 50-year life plan, a 30-year vision in 2010, a company meant to last 300 years.

Jan 1975At 17 he saw a photo of a microchip that, by his account, set his direction for 50 years.

Jun 2026Set a target of ¥1 quadrillion in net asset value by 2042, about 14 times the June 2026 figure.

“A smooth succession of the business so that SBG can continue to grow for 300 years.”Annual Report 2021

Build a group, not a company

He runs SoftBank as a cluster of number-one companies tied together by capital, not as one operating business.

2017Raised the $100B Vision Fund, $45B of it from Saudi Arabia’s Public Investment Fund.

Apr 2026Made Arm chief Rene Haas CEO of SoftBank Group International, over its chip and AI holdings.

“I am not a CEO. I am building an empire.”To an investor in 2020, per Barber’s Gambling Man

Decide fast, on conviction

He commits on his reading of a founder and a trend, sometimes in one meeting, and the misses are as large as the wins.

2000Paid $20M for an Alibaba stake after a six-minute meeting with Jack Ma; worth about $150B by 2020.

2019WeWork, valued at $47B on his backing, cost SoftBank about $11.5B in equity. He called it a stain.

“I firmly believe that OpenAI is on track to become the most valuable company on the planet.”SoftBank Group Report 2025

Announce the big number first

He names a very large commitment in public, often beside a head of state, then goes and raises the money.

Dec 2024Pledged $100B of US investment over four years, standing beside Donald Trump at Mar-a-Lago.

Jan 2025Became chairman of Stargate, announced at the White House with a $500B target by 2029.

“President Trump is a double-down president, I’m going to have to double down.”Mar-a-Lago, December 2024

Borrow against the winners

He pays for each bet with debt and asset sales backed by earlier wins, inside a loan-to-value ceiling of 25%.

Oct 2025Sold 32.1M Nvidia shares for $5.8B to help fund the OpenAI commitment.

Mar 2026Signed a $40B unsecured one-year bridge loan with five banks; prepaid the balance in September.

“The greater the risk capital providers take to shape the future, the greater the expected return.”Annual Report 2021

Own the layer under everything

He wants the chip design that every device and data centre runs on, then the chips, power and robots around it.

Sep 2016Completed the £24B purchase of Arm, then listed it in September 2023 and kept control.

Nov 2025Closed the $6.5B purchase of chip designer Ampere, a month after agreeing to buy ABB Robotics for $5.4B.

“Arm has overwhelming strength in semiconductor design intellectual property.”SoftBank Group Report 2026

Admit the loss, then attack

After a failure he says so in public, cuts back and holds cash, then returns to offence when the footing allows.

Nov 2020Told the DealBook conference he held almost $80B in cash to be ready for a worst case.

Aug 2022Reported a ¥3.16T quarterly loss, SoftBank’s largest, and said he was ashamed of his earlier pride.

“We now have the solid financial footing we need to play offense.”SoftBank Group Report 2023

Evolution

How the thinking changed

YearsEraWhat they believedWhat changed it

1981–1994

PC software

The PC era needs a software distributor at its centre.

Founded SoftBank on 3 September 1981 at 24 and listed it in July 1994 at a market value of about ¥200B.

1995–2000

Internet

Own a piece of the internet’s leaders in every market.

Backed Yahoo, started Yahoo Japan in 1996 and bought into Alibaba in 2000. The crash then cut SoftBank’s value by 98%.

2001–2013

Broadband and mobile

An internet company has to own the network.

Launched broadband with Yahoo Japan in 2001, bought Vodafone Japan in 2006, brought in the iPhone in 2008 and Sprint in 2013.

2014–2019

Arm and Vision Fund

Capital at scale can pick the winners of the AI age.

Bought Arm for £24B and raised a $100B fund, then saw WeWork collapse from a $47B valuation in 2019.

2020–2022

Defence

Sell assets, hold cash and buy back shares.

A record ¥4.99T profit in the year to March 2021, then record Vision Fund losses and a ¥3.16T quarterly loss in 2022.

2023–2026

ASI offence

Superintelligence comes in about ten years; SoftBank must be its platform.

Arm’s listing, Stargate, $64.6B for OpenAI, Ampere, ABB Robotics and DigitalBridge; a record ¥5.0T profit to March 2026.

Scale

The numbers behind the posture

$20M

Paid for an Alibaba stake in 2000, worth about $150B by 2020

−98%

Fall in SoftBank’s market value after its $180B peak in 2000

$47B

WeWork’s valuation in early 2019; it cost SoftBank $11.5B in equity

¥3.16T

Net loss for April to June 2022, SoftBank’s largest quarterly loss

$64.6B

Cumulative investment in OpenAI by October 2026, for about 13%

¥5.0T

Net income for the year to March 2026, SoftBank’s highest ever

Tensions

Where the model strains

A company meant to grow for 300 years

Its market value fell 98% after 2000, and it lost ¥3.16T in one quarter of 2022

A life plan that hands over the business in his 60s

At 68 he rewrote the plan to keep going into his 70s; he is still chairman and CEO

Net asset value of ¥74T, by his own June 2026 slide

A market value of ¥37T on the same slide, half of what he says SoftBank is worth

“Some people say that AI may be a bubble. I disagree.”

The same conviction about the internet cost him about $70B of personal wealth in 2000

For operators

What to take from it

Pick the next platform early and move real capital onto it, not a pilot.

Keep one liquid winner you can sell or borrow against when a bet fails.

Write the long plan down with dates, so people know what you will do next.

When a bet fails, say so in public and tighten your own rules before the next one.

Sources

SoftBank Group Reports and CEO messages (2021, 2023, 2024, 2025, 2026) · Message from the Chairman & CEO, group.softbank (2026) · Annual General Meeting presentations (June 2024, 2025, 2026) · SoftBank earnings presentations (November 2019, August 2022, May and August 2026) · SoftBank press releases (February, March, April and September 2026) · SoftBank Annual Report 2010 · TechCrunch (2020, 2024, 2025, 2026) · Lionel Barber, Gambling Man (2024), via TechCrunch · New York Times DealBook (2020), via TechCrunch · Wikipedia entries on Son, SoftBank, Sprint, Stargate and the Vision Fund, with reports they cite from the Wall Street Journal (2000), the Guardian and CNBC (2022). SoftBank documents are quoted from the company’s English versions of Japanese originals. Quotes shortened.

All profilesRuchi Sanghvi

Founder of South Park Commons · 2005 to 2026

Ruchi
Sanghvi

The Minus One Architect

She treats the time before a founder has an idea as a stage with its own craft: explore full time, among peers, and discard good ideas until one is worth a decade. South Park Commons is that belief turned into an institution.

Drawn from her essay “What is -1 to 0?” (2023, revised 2026), her South Park Commons 10-year note (2026), SPC’s fund announcements (2025 and 2026), a Carnegie Mellon profile (2010), and reporting by TechCrunch, AllThingsD and Forbes (2006 to 2025).

The central idea

The road from -1 to 0

Before 0 to 1 there is a stage where the founder has no idea yet. She treats it as a craft with its own method, and built South Park Commons to house it.

Swipe sideways to see the full diagram

At -1The SquiggleEverything you could do and no conviction yet. Most people, she writes, do not sit in it for long.

→

-1 to 0Ideation as a jobExplore full time among peers at the same stage. Discard good ideas in search of a great one.

→

At 0ConvictionOne problem worth the next decade. She calls it founder-market fit.

→

0 to 1BuildOnly now raise money and search for product-market fit.

6 monthsMember residency, at no cost or equity

$400K for 7%Founder Fellowship, plus $600K in the next round

$575MFund IV, August 2026

1,200Members across three cities, August 2026

Stages as described in her essay “What is -1 to 0?” (2023, revised August 2026). Program terms are from the South Park Commons site and its Fall 2026 fellowship notice; fund and member figures are from the Fund IV announcement of 5 August 2026.

Operating system

Eight principles

Start before the idea

Choosing the problem is its own stage, and it comes before 0 to 1.

2015–16Gathered ten people around her dining table with no plan; it became South Park Commons.

Aug 2026Republished her 2023 essay defining -1 to 0 as the stage before a company exists.

“What problem is worth the next decade of your life.”“What is -1 to 0?”, SPC blog, 2026 (shortened)

Treat ideation like a job

Conviction comes from full-time, structured exploration, not from waiting for inspiration.

2018Members paid $600 for six months of full-time exploration at SPC, with some fees waived.

2026The member residency runs six months and takes no fee and no equity.

“You will not have a eureka moment while meditating on top of a mountain.”“What is -1 to 0?”, SPC blog, 2026

Discard good for great

A good idea with early traction can be a trap. Keep looking for the one worth ten years.

2021Render, begun at SPC by Stripe’s eighth employee after he set aside an AI idea, raised a $20M Series A.

Aug 2026Wrote that it is now easier than ever for a little traction to lure a founder down a dead end.

“You should be more afraid of local maxima.”“What is -1 to 0?”, SPC blog, 2026

Go slow to go fast

Time spent on conviction up front saves years on the wrong company.

2011–12By her account, the year after she left Facebook to start a company felt like hitting a wall.

2021Members typically stayed about nine months before leaving to build, TechCrunch reported.

“Spend a bit of it upfront so you don’t waste more of it later.”“What is -1 to 0?”, SPC blog, 2026

Do not explore alone

Peers at the same uncertain stage test ideas and make the uncertainty bearable.

2021About half of members found co-founders or early hires inside the community, TechCrunch reported.

Jan 2026Reported 25,000 applications a year and more than 1,000 members across three cities.

“Where ambition didn’t require loneliness.”SPC 10-year anniversary post, 2026

Back the founder first

Fund the person before the idea is fixed. The fit to look for is founder to market.

2018Raised a first fund to back members, reported at $40M that April and later put at $55M.

2026Fellowship terms: $400K for 7% on a SAFE, plus $600K guaranteed in the next round.

“-1 to 0 is about Founder-Market Fit.”“What is -1 to 0?”, SPC blog, 2026

Build what you use daily

She joined a product she already used, then shipped its most contested feature.

2005Left Oracle for Facebook, then about a year old with a staff of 15 to 20.

Sep 2006Wrote the News Feed launch post; about 100,000 users joined a protest group within a day.

“Working on a product that I used every day and was passionate about clinched the deal.”Carnegie Mellon profile, 2010 (shortened)

Be heard in the room

Facebook’s first female engineer, she describes her own style as aggressive.

2011Won a TechFellow award for engineering leadership for her work at Facebook.

2011A HuffPost headline quoted her: “It Was Difficult To Break Into The Boys’ Club.”

“I just happen to be a woman who is aggressive.”Mint interview headline, 2013

Evolution

How the thinking changed

YearsEraWhat they believedWhat changed it

2005–2010

Facebook

Work on the product you use every day.

News Feed drew a user revolt in September 2006 and stayed. She later led product for Platform and Connect.

2011–2012

Cove

Leave the rocket ship and start a company.

A year that felt like “banging my head against a wall.” Dropbox bought Cove, still in stealth, in February 2012.

2012–2013

Dropbox

Scale a company by recruiting.

Headcount went from 90 to 265. She left as VP of Operations in October 2013, citing a “breakneck pace.”

2014–2017

The kitchen table

The gap between companies needs a place and peers.

After Facebook, Cove and Dropbox came, in her words, “nothing.” Ten people at her table became SPC in 2016.

2018–2023

Community with a fund

Back members when they reach conviction.

A $55M first fund, Aditya Agarwal joining in 2018, a $150M second fund and a New York hub from 2021.

2024–2026

Three cities

-1 to 0 is a stage worth an institution.

Bengaluru opened in 2024. A $275M fund in 2025 and a $575M fund in August 2026 took assets to $2B.

Scale

The numbers behind the posture

$575M

Fund IV, announced August 2026, SPC’s largest

$2B

Assets under management across SPC funds, August 2026

1,200

Members in the SPC network, August 2026

25,000

Applications a year, by her count in January 2026

~100K

Users in a News Feed protest group within a day, September 2006

265

Dropbox staff when she left in October 2013, up from 90

Tensions

Where the model strains

“Go slow to go fast”

Her own startup, Cove, was sold to Dropbox in about a year, still in stealth

The residency takes no fee and no equity

The firm around it manages $2B (2026) and takes 7% from fellows

“Where ambition didn’t require loneliness”

25,000 apply each year; the whole network is about 1,200 people

Says Facebook “made the web more social”

Users revolted at News Feed in 2006; Zuckerberg wrote “We really messed this one up”

For operators

What to take from it

Treat choosing the problem as a job, with hours, peers and a deadline.

Drop a good idea early if you would not give it ten years.

Do your exploring in a room of people at the same stage.

Raise money once you have conviction, not in order to find it.

Sources

Ruchi Sanghvi, “What is -1 to 0?” (SPC blog, 2023, revised August 2026) · Ruchi Sanghvi, “South Park Commons 10-Year Anniversary” (January 2026) · SPC Fund III (May 2025) and Fund IV (August 2026) announcements · SPC Founder Fellowship notice (Fall 2026) · southparkcommons.com (2026) · Carnegie Mellon University profile (2010) · TechCrunch (2006, 2012, 2013, 2021, 2025) · AllThingsD (2013) · Forbes (2018) · Mint (2013) and HuffPost (2011) headlines via Wikipedia · Wikipedia entries on Sanghvi, South Park Commons and Facebook’s Feed. Quotes shortened.

All profilesSean Parker

Napster co-founder, first Facebook president · 1999 to 2026

Sean
Parker

The Transition Hacker

He treats the end state as obvious and the route as the work: prove the demand, license it later, keep the founder in control. Then he turned the same hacker’s eye on philanthropy and on the products he built.

Drawn from his Axios interview (2017), his Wall Street Journal essay “Philanthropy for Hackers” (2015), his Forbes cover profile (2011), his TechCrunch essay on his wedding (2013), the Daily Beast summit (2010) and reporting from David Kirkpatrick’s The Facebook Effect (2010).

The central idea

The transition is the strategy

Napster proved people wanted every song on demand, and the courts shut it down. A decade later he backed the licensed version and plugged it into Facebook.

Swipe sideways to see the full diagram

1999 · No licencesNapsterLaunched in June 1999. 26.4M verified users by February 2001. Offline by court order in July 2001.

→

2010 · LicensedSpotifyHe invested, sat on the board and helped win over Warner and Universal for the US launch.

→

2011 · DistributionFacebookSpotify reached the US in July. In September it was built into Facebook at the f8 conference.

→

2026 · Scale777M listenersSpotify’s monthly active users in June 2026. He left its board in June 2017.

The same route again in AI: executive chairman of Stability AI since June 2024. Universal, Sony and Warner Music joined its 2026 Series B. $76M

Napster figures are its verified peak in February 2001 and its shutdown in July 2001. Spotify users are for June 2026. The $76M is the size of Stability AI’s 2026 Series B.

Operating system

Eight principles

Work on the transition

Most people agree where things are heading. The advantage is in knowing how to get there.

2009Emailed Spotify’s Daniel Ek: “You’ve built an amazing experience.” He invested in 2010.

Sep 2011Announced Spotify’s integration with Facebook at the f8 conference.

“The transition strategies are more important than understanding what the outcome state will be.”Forbes, 2011

Convenience beats enforcement

People pirate because it is easier. The answer is a legal product that is more convenient.

Oct 2010Put illegal downloads at 4 to 10 trillion songs, against 4 billion bought legally.

2010Spotify then had 10 million users in Europe and over 500,000 paying subscribers.

“You have to accept that the war on piracy is a failure.”Daily Beast summit, via CBS News, 2010

Failure is tuition

He counts Napster as his schooling in copyright law, corporate finance and what not to do.

Dec 1999The record labels sued Napster for copyright infringement six months after launch.

Sep 2001Napster agreed to pay $26M for past use and a $10M advance on future royalties.

“So much of what I learned at Napster was learning what not to do.”Forbes, 2011

Keep the founder in control

Pushed out of Plaxo by its investors, he set up Facebook so Zuckerberg could not be removed.

2004Brought in Peter Thiel, whose $500,000 for 10.2% was Facebook’s first outside money.

Apr 2005Accel put in $12.7M at a $98M valuation; Zuckerberg controlled three of five board seats.

“Sean was pretty material in setting up the company in a way that Mark retained as much control as he does.”Dustin Moskovitz, in Forbes, 2011

Build the spread into the product

Growth is designed in, not added later. Critics called Plaxo’s update emails spam.

Nov 2002Launched Plaxo, an address book that grew by emailing each user’s contacts for updates.

2007–13Causes, an early Facebook app, reached 186M users and $50M in donations by 2013.

“Plaxo is like the indie band that the public doesn’t know but was really influential.”Forbes, 2011

Say what you built

He said in public that Facebook was designed to take attention, and that its makers knew.

Nov 2017Told an Axios event its builders “understood this consciously. And we did it anyway.”

Dec 2017Facebook answered publicly after a second former executive made similar claims.

“You’re exploiting a vulnerability in human psychology.”Axios, November 2017

Give like a hacker

Large, concentrated bets on problems that look solvable, joining money, science and policy.

Jun 2015Put $600M into the Parker Foundation for life sciences, public health and civic engagement.

Apr 2016Gave $250M to start a cancer immunotherapy institute across six research centres.

Hackers share an antiestablishment bias, a nose for weak points in systems and a faith in data.“Philanthropy for Hackers”, WSJ, 2015 (paraphrase)

Bring the rights holders inside

The Napster lesson applied to generative AI: make the owners of the content your partners.

Jun 2024Became executive chairman of Stability AI in a rescue round of about $80M.

2026Universal, Sony, Warner Music and Electronic Arts backed Stability’s $76M Series B.

“Innovation happens at the intersection of art and technology.”Stability AI statement, via TechCrunch, 2024

Evolution

How the thinking changed

YearsEraWhat they believedWhat changed it

1996–2001

Hacker and Napster

Music should move freely online; the industry will have to adapt.

The labels sued in 1999. He was sent on an “extended vacation” that became his exit; Napster went offline in July 2001.

2002–2004

Plaxo

A product that spreads itself needs no marketing.

Investors removed him in 2004. His co-founders said he was often absent; he says he was forced out.

2004–2005

Facebook president

A founder needs structural control to survive his investors.

Police found cocaine at a house he rented in August 2005. He was not charged, and resigned under investor pressure.

2006–2014

Investor and Spotify

The licensed version of Napster wins if it is easier than piracy.

Founders Fund partner from 2006. Spotify reached the US in 2011. Airtime had a glitchy debut in 2012. He left the fund in 2014.

2015–2023

Hacker philanthropist

Philanthropy should make big bets on solvable problems and get political.

$600M foundation, $250M cancer institute, Opportunity Zones in the 2017 tax law. His civic startup Brigade broke up in 2019.

2024–2026

AI and cancer vaccines

Generative AI needs the owners of the content as partners.

Chairs Stability AI; music labels invested in 2026. His institute pushed cancer vaccines; firms it seeded sold to Lilly and BMS.

Scale

The numbers behind the posture

26.4M

Napster’s verified peak users, February 2001

3 of 5

Facebook board seats Zuckerberg controlled after the April 2005 round

777M

Spotify monthly active users, June 2026; he invested in 2010

$250M

Parker Institute grant, April 2016; $600M to his foundation, 2015

$2.5M

Settlement with California’s Coastal Commission over his 2013 wedding

$3.4B

Forbes estimate of his net worth, 1 October 2026

Tensions

Where the model strains

Says Facebook exploits “a vulnerability in human psychology”

Facebook is the source of his $3.4B fortune, by Forbes’s 2026 estimate

Said in 2011 he pays “far too little” tax, above all on capital gains

Drove Opportunity Zones, a capital gains tax break, into the 2017 tax law

Founders should be protected from their investors

Forbes called the result “complete and permanent control” for Zuckerberg

Said he would not “run roughshod” over the redwood grove where he married

No coastal permit for the set; paid $2.5M, by his account the commission’s largest

For operators

What to take from it

Assume the end state is obvious. Spend your effort on the route to it.

If demand is proven outside the rules, build the licensed version that is easier.

Settle control in the documents before the first big round, not after.

Know what your product does to its users, and say it before your critics do.

Sources

Axios (2017) · Forbes (2011, 2026) · Sean Parker, The Wall Street Journal (2015) · Sean Parker, TechCrunch (2013) · CBS News (2010) · The Next Web (2011) · San Francisco Bay Guardian (2011) · TechCrunch (2012 to 2024) · David Kirkpatrick, The Facebook Effect (2010) · SFGate (2014) · PR Newswire (2016) · Parker Institute for Cancer Immunotherapy (2026) · Wikipedia entries on Napster, Facebook, Spotify and Stability AI (October 2026). Quotes shortened; paraphrases marked.

All profilesPeter Thiel

Co-founder of PayPal and Palantir; investor · 1998 to 2026

Peter
Thiel

The Contrarian Monopolist

He asks what important truth few people agree with him on, then backs founders who can turn that secret into a monopoly. He applies the same method to politics and religion, where his conclusions are far more disputed.

Drawn from Zero to One (2014), “The Education of a Libertarian” (2009), “The Straussian Moment” (2007), the Founders Fund manifesto (2011), Conversations with Tyler (2015, 2024), Uncommon Knowledge (2024), Interesting Times and his First Things essay (2025).

The central idea

Escape competition, build a monopoly

Copying what works takes the world from 1 to n and competes the profit away. Making something new goes from 0 to 1, and its maker can own the market. His companies and his fund follow the rule.

Swipe sideways to see the full diagram

1 to nCompetitionCopy what already works and add more of something familiar.Rivals who look alike fight over the same customers, and the profits are competed away.“In perfect competition, a business is so focused on today’s margins that it can’t possibly plan for a long-term future.”

0 to 1Creative monopolyStart from a secret: an important truth that very few people agree with.Make something new, own its market, and aim to be the last mover in it.“Monopolists can afford to think about things other than making money; non-monopolists can’t.”

$1.5BPayPal, sold to eBay in October 2002

$500KFor 10.2% of Facebook in 2004; sold for over $1B

$1.94BPalantir revenue in Q2 2026, up 93%

$6BFounders Fund Growth IV, as reported in March 2026

Quotations from Zero to One (2014). Figures: eBay’s purchase of PayPal, October 2002; his Facebook investment of August 2004 and his share sales of $638M and $395.8M in 2012; Palantir’s results for the quarter to 30 June 2026; Founders Fund Growth IV as reported by TechCrunch in March 2026.

Operating system

Eight principles

Ask the contrarian question

He looks for an important truth that few people accept, then builds or invests on it.

Aug 2004Put $500,000 into Facebook for 10.2%, the first outside money in a site for college students.

2008Founders Fund put $20 million into SpaceX, then an unproven private rocket company.

“What important truth do very few people agree with you on?”Zero to One, 2014

Competition is for losers

Profit comes from being the only one doing something, not from beating rivals at the same thing.

Mar 2000Merged Confinity with its rival X.com rather than compete; the combined company became PayPal.

Jul 2025The US Army gave Palantir a single enterprise agreement worth up to $10 billion over 10 years.

“The goal of every successful business is to have a monopoly.”Conversations with Tyler, 2015

Returns follow a power law

A few investments outweigh all the rest, so he concentrates money in the ones that could.

2012Sold most of his Facebook stake for over $1 billion, on $500,000 invested eight years earlier.

Jun 2025Founders Fund wrote its largest cheque, $1 billion, into Anduril at a $30.5 billion valuation.

The best investment in a successful fund equals or outperforms the rest of the fund combined.Zero to One, 2014 (paraphrase)

Long substance, short status

He treats elite credentials as a status contest and pays talented people to skip them.

Sep 2010Announced the Thiel Fellowship: $100,000 over two years to leave college and build something.

2026The grant is now $250,000; past fellows co-founded Figma, Ethereum and Scale AI.

“I would always be long substance, short status.”Conversations with Tyler, 2015

Bits moved, atoms stalled

He argues that progress since the 1970s has been fast in computing and slow in the physical world.

2011Founders Fund’s manifesto, “What Happened to the Future?”, committed it to harder technology.

2025Backed space factory Varda, parts maker Hadrian and uranium enricher General Matter.

“We wanted flying cars, instead we got 140 characters.”Founders Fund manifesto, 2011

Exit over voice

His 2009 essay said libertarians should escape politics through technology rather than win votes.

2009Wrote that he no longer believed “that freedom and democracy are compatible.”

2026Left California before a proposed tax on billionaires, for Miami and then Buenos Aires.

“The great task for libertarians is to find an escape from politics in all its forms.”The Education of a Libertarian, 2009

Fund what peers will not

He puts money and his name behind causes most of Silicon Valley avoids.

2016Paid about $10 million for Hulk Hogan’s suit against Gawker, then spoke for Trump at the convention.

2021–22Gave $15 million to a super PAC behind J.D. Vance’s Senate run in Ohio.

“I only want to fund unpopular causes.”Conversations with Tyler, 2015

Fear the cure as well as the risk

He argues that a world state built to prevent catastrophe is as dangerous as the catastrophe.

Sep 2025Began four off-the-record lectures on the Antichrist in San Francisco; the audio leaked in October.

Mar 2026Repeated the series in Rome; a priest who advises the Vatican on technology called it a “radicalization.”

“I don’t want Antichrist, I don’t want Armageddon.”Uncommon Knowledge, 2024

Evolution

How the thinking changed

YearsEraWhat they believedWhat changed it

1985–1998

Stanford and law

The elite track is a contest worth winning.

Co-founded The Stanford Review in 1987, took up René Girard’s theory of imitation, and finished Stanford Law in 1992.

1998–2002

PayPal

Digital money can free people from weak currencies.

Took PayPal public in February 2002 and sold it to eBay that October for $1.5 billion; his 3.7% was worth $55 million.

2002–2010

Macro and Facebook

Big macro calls and early startup bets can both work.

Clarium peaked above $7 billion in 2008 and shrank to about $350 million by 2011. Facebook and Palantir were the winners.

2010–2015

Teacher of startups

Technology has stalled, and founders with secrets can restart it.

The Fellowship, the manifesto, a 2012 Stanford class and Zero to One in 2014 made him a public thinker.

2016–2022

Political turn

The government is broken, so back outsiders.

Funded the Gawker suit, spoke for Trump, spent over $35 million on the 2022 midterms and left Meta’s board.

2023–2026

Allocator and lecturer

AI is real but bubbly; a one-world state is the larger risk.

Said he would sit out 2024 as Vance became vice president; sold Nvidia, lectured on the Antichrist, moved to Buenos Aires.

Scale

The numbers behind the posture

$500K

Paid for 10.2% of Facebook in August 2004

$1.94B

Palantir revenue in Q2 2026, up 93% on a year earlier

$6B

Founders Fund Growth IV, reported March 2026; ~$1.5B from partners

$36.6B

His net worth on 29 September 2026, per Forbes

$15M

Given to a super PAC for J.D. Vance’s Ohio Senate race, 2021 to 2022

$5B

His Roth IRA by 2019, grown tax-free from $1,700, per ProPublica

Tensions

Where the model strains

Urged “an escape from politics in all its forms” in 2009

Spent over $35 million on the 2022 midterms; a former employee he funded is vice president

Says a totalitarian one-world state is a risk as grave as any other

Palantir, which he chairs, sells data tools to ICE and the Army; critics see surveillance

Calls funding the Gawker suit deterrence, and philanthropy

It was funded in secret and bankrupted a publisher; critics warned of a chilling effect

Says the problem is stagnation: too little technological progress

Max Chafkin’s The Contrarian (2021) says he pursues it with little regard for the costs

For operators

What to take from it

Write down what you believe that few others do, then test whether it is true.

Pick a market small enough to own before you try to scale.

Size bets for a power law: one winner should be able to carry the fund.

Think for yourself. Opposing the crowd by reflex is just another way to follow it.

Sources

Peter Thiel with Blake Masters, Zero to One (2014) · “The Education of a Libertarian,” Cato Unbound (2009) · “The Straussian Moment” (2007) · Founders Fund manifesto (2011) · Conversations with Tyler (2015, 2024) · Uncommon Knowledge, Hoover Institution (2024) · The Joe Rogan Experience #2190 (2024) · Republican convention speech via Time (2016) · Interesting Times, The New York Times (2025), via Fortune · Thiel and Wolfe, First Things (2025) · Max Chafkin, The Contrarian (2021), via Kirkus · ProPublica (2021) and The New York Times (2016), via Wikipedia · Forbes (2026) · Fortune (2025, 2026) · TechCrunch (2024, 2025, 2026) · thielfellowship.org · Farnam Street. Quotes shortened; paraphrases marked.

How Founders Think

MarkZuckerberg

The Dictator’s Dilemma

Eight principles

Think in empires

Never rent the platform

Buy or copy before it matures

Speed first, then stability

Control is the moat

Hoard signal, fight the filter

Reset hard when behind

Building is identity

How the thinking changed

The Hacker

Mobile survival

Crisis and defense

The metaverse

Efficiency and open AI

Superintelligence

The numbers behind the posture

Where the model strains

What to take from it

JensenHuang

The long fuse

Eight principles

Speed of light

Zero-billion-dollar markets

Pay for conviction

One room, no one-on-ones

Read the weak signals

Suffering as training

Admit the wrong bet, fast

Obsolete yourself on schedule

How the thinking changed

The graphics bet

The GPU as a processor

Accelerated computing

All-in on deep learning

The data center company

AI factories

The numbers behind the posture

Where the model strains

What to take from it

ElonMusk

The Algorithm

Question

Delete

Simplify

Accelerate

Automate

Eight principles

First principles, not analogy

Delete before you optimize

Physics is the law

Maniacal urgency

The factory is the product

Hardcore by default

Bet everything, twice

Truth-seeking as the mission

How the thinking changed

Internet builder

Physics contrarian

Reuse and scale

Manufacturing is hard

Cut hard, build compute

State, then space compute

The numbers behind the posture

Where the model strains

What to take from it

BarryDiller

Build, operate, spin, repeat

Eight principles

Start from a blank page

Creative conflict

Hire the discontented

Instinct over research

Own, don’t rent

Walk away on price

Spin-offs as renewal

Don’t bet the farm

Mark
Zuckerberg

Jensen
Huang

Elon
Musk

Barry
Diller

Arthur
Blank

Amjad
Masad

Andy
Jassy