../

Other founders

What operators, not investors, actually said about building companies: Jeff Bezos, Reid Hoffman, Brian Chesky, the Collisons, Jason Fried and DHH, Tobi Lütke, Stewart Butterfield, Sara Blakely and Jensen Huang, each with the source and where the idea applies in the idea-to-MVP playbook. The investor views are in Y Combinator, a16z and Peter Thiel; Elon Musk, Steve Jobs and Jensen Huang have their own sheets (Huang gets a short summary here).

Who's here and why

FounderCompanyBest known for (as advice)Main source
Jeff BezosAmazonDay 1, customer obsession, reversible decisions, memosannual shareholder letters (1997–2020)
Reid HoffmanLinkedIn (co-founder), PayPal (executive)launch while embarrassed; blitzscalingLinkedIn essays, Blitzscaling (2018), Masters of Scale podcast
Brian CheskyAirbnbhandcrafted experiences, the 11-star exerciseMasters of Scale (2017), Stanford CS183B (2014)
Patrick & John CollisonStripeaggressive manual onboarding; speedPaul Graham's essays, patrickcollison.com
Jason Fried & David Heinemeier Hansson (DHH)37signals / Basecampbuild less, fund yourself, calm growthGetting Real (2006), Rework (2010), It Doesn't Have to Be Crazy at Work (2018)
Tobi LütkeShopifythe trust batteryThe Knowledge Project podcast (2018)
Stewart ButterfieldSlacksell the outcome, not the tool"We Don't Sell Saddles Here" (memo 2013, published 2014)
Sara BlakelySpanxfailure as not tryingessay in Getting There (2015)
Jensen HuangNvidiafirst principles; resilience through hardshipStanford GSB (2024)

Founders' advice has the opposite bias to investors': it comes from one company, told in hindsight. Treat each as a tool, not a law.

Jeff Bezos (Amazon)

Bezos's annual letters to shareholders are the richest primary source any founder has left. The 1997 letter was re-attached to every later letter.

PrincipleWhat it meansSourceApply it by…
It's all about the long termmeasure success by long-term shareholder value; invest for market leadership over short-term profits; "we'll take the cash flows" over GAAP appearances; make bold rather than timid bets1997 letter (opens in a new tab) (attached to later letters)deciding up front what you'll optimize for
Obsess over customershe says obsessive customer focus is by far the most protective of "Day 1" vitality, because customers are "always beautifully, wonderfully dissatisfied"2016 letter (opens in a new tab)customer discovery that never stops
Day 1 vs Day 2"Day 2 is stasis. Followed by irrelevance. Followed by excruciating, painful decline. Followed by death." His Day 1 defenses: customer obsession, a skeptical view of proxies, eager adoption of external trends, high-velocity decisions2016 lettera startup is Day 1 by default; the risk begins once process appears
Resist proxiesprocess becomes the thing ("we followed the process"); surveys become a proxy for customers2016 letterwatching real users rather than only survey averages
Type 1 / Type 2 decisionsone-way doors (irreversible) deserve slow, careful thought; two-way doors (reversible) should be made quickly by individuals or small groups; as organizations grow they tend to use the heavy process for most decisions, including Type 2 ones2015 letter (opens in a new tab)decision rules: classify each decision's reversibility
Decide at ~70%"most decisions should probably be made with somewhere around 70% of the information you wish you had"; waiting for 90% is usually slow; being good at course-correcting makes being wrong cheap2016 lettera deadline on each open decision
Disagree and commitsay "Look, I know we disagree on this but will you gamble with me on it?" He does it too, and says escalate true misalignment early instead of deciding by exhaustion2016 letterco-founder disputes on two-way doors
Failure and invention are twinshe calls Amazon "the best place in the world to fail"; "Given a ten percent chance of a 100 times payoff, you should take that bet every time." Unlike baseball, business outcomes are long-tailed2015 letterrunning several cheap experiments in validation
High standardsteachable, not innate; domain-specific; you must recognize what good looks like; and set realistic scope (his handstand example: people expect two weeks; it takes six months)2017 letter (opens in a new tab)estimating honestly how long "good" takes
Six-page narrativesno slide decks; narratively structured six-page memos, read silently at the start of meetings; great memos take a week or more of rewriting2017 letterwriting the one-page spec in prose
Wanderingefficiency when you know where you're going; guided "wandering" for non-linear discoveries2018 letter (opens in a new tab)protecting some unplanned exploration

Working backwards and the PR/FAQ

Colin Bryar and Bill Carr, both long-time Amazon executives, describe Amazon's processes in Working Backwards (St. Martin's Press, 2021). The core ones, as they describe them:

MechanismWhat it isApply it by…
PR/FAQbefore building, write the launch press release (short, customer-facing) plus FAQs for customers and internal stakeholders; if the press release isn't exciting, the product isn't worth buildingwriting a PR/FAQ for your MVP in product design
Narrativessix-page memos instead of slides (as in the 2017 letter)decisions argued in prose
Bar Raisera trained interviewer from outside the team with veto power in hiringan outside voice in early hires
Single-threaded leadersone person whose only job is one initiative; the book describes this evolving from the earlier "two-pizza team" idea (small teams two pizzas can feed)one owner per area
Input metricsmanage the controllable inputs (selection, price, speed) rather than output metrics like revenueinstrumentation of leading indicators

Regret minimization

In a 2001 interview for the Academy of Achievement (opens in a new tab), Bezos described deciding to leave his Wall Street job in 1994 with what he called a "regret minimization framework": project yourself to age 80 and minimize the number of regrets. "I knew that when I was 80 I was not going to regret having tried this." He adds that most regrets are "acts of omission". Use it for the decision to go full-time.

Reid Hoffman (LinkedIn)

Launch while embarrassed

Hoffman coined the line, by his own account more than a decade before his 2017 essay If There Aren't Any Typos In This Essay, We Launched Too Late! (opens in a new tab):

"If you're not embarrassed by the first version of your product, you've launched too late."

His three themesWhat he means
speedaim to launch fast
your assumptions are wrongyou'll be embarrassed by how many wrong assumptions you made about customers
feedback looplaunching late delays learning

His limits on it: "I didn't write 'If you're not indicted by the first version of your product…'". If a launch generates lawsuits, alienates users or burns resources for no gain, it was too soon. Eric Ries's corollary, which he quotes: however long you wait, you'll be embarrassed anyway. His example: LinkedIn launched in 2003 without "Contact Finder", a feature the team thought essential; 14 years later it still didn't exist. Apply it to what not to build.

Blitzscaling (2018)

Blitzscaling (Reid Hoffman and Chris Yeh, Currency, 2018) is about the next stage: once you have product–market fit in a winner-take-most market, deliberately prioritize speed over efficiency, accepting waste and chaos to win first.

Stage (book's names)Rough employee count
Family1–9
Tribetens
Villagehundreds
Citythousands
Nationtens of thousands

This is not MVP advice. Blitzscaling before PMF is just burning money faster; see PG's default alive.

Masters of Scale: "handcrafted"

Hoffman's podcast opened in 2017 with an episode on Chesky called "Handcrafted", arguing that to scale you first have to do things that don't scale (transcript via Tim Ferriss #326 (opens in a new tab)). His summary: dream big, act small, handcraft the core service, then work out which part of it can scale.

Brian Chesky (Airbnb)

IdeaWhat Chesky saidSourceApply it by…
Live with your users"when you bought an iPhone Steve Jobs didn't come sleep on your couch, but I did"; features like profiles and reviews came from staying with hostsMasters of Scale, "Handcrafted" (2017)customer discovery in person
The 11-star exercisetake one part of the product, describe the 5-star version, then 6, 7… up to an absurd 11 (a Beatles-style welcome, a trip to space); the sweet spot for design is somewhere between "they opened the door" and "I went to space"same; Hoffman's write-up (2018) (opens in a new tab)designing the core experience of the MVP
Two stages"There's really two stages of a startup's product. The first is design a perfect experience and then you scale that experience.""Handcrafted" (2017)perfecting the experience for a few users before growth
Cereal boxesbroke and in debt in 2008, the founders made and sold election-themed cereal (Obama O's, Cap'n McCain's); he says they made $40,000 selling cereal that year and $5,000 from the websiteHow to Start a Startup, Stanford CS183B, lecture on culture (2014) (opens in a new tab)relentless resourcefulness to buy runway
Founder modehis 2024 YC talk on how the "hire good people and give them room" advice failed him prompted PG's essayFounder Mode (opens in a new tab) (PG, 2024)staying in the detail as you grow

The 11-star exercise and "live with your users" are Chesky's versions of PG's Do Things That Don't Scale (opens in a new tab) (2013), which uses Airbnb's door-to-door work in New York as its lead example.

Patrick and John Collison (Stripe)

IdeaWhat happenedSourceApply it by…
The Collison installationYC's term for the brothers' habit: when someone agreed to try Stripe they'd say "Right then, give me your laptop" and set it up on the spotPG, Do Things That Don't Scale (opens in a new tab) (2013)onboarding each early user yourself, immediately
Be the softwareStripe's "instant" merchant accounts were set up manually by the founders behind the scenessamea concierge MVP
A tiny MVPSeibel describes Stripe's first version (then called /dev/payments) as having almost no features, a small bank partner and nightly manual paperwork, aimed only at early YC startupsSeibel, How to Build an MVP (opens in a new tab) (2023)aiming v1 at one narrow, desperate segment
Support at 3 a.m.Patrick describes a bot that paged a founder if a chat-room question went unanswered for 30 secondsMasters of Scale, "Handcrafted" (2017)founder-run support early on
FastPatrick's page listing ambitious projects done quickly (e.g. the Empire State Building in 410 days, Apollo 8 decided and launched in 134 days, JavaScript's prototype in 10 days)patrickcollison.com/fast (opens in a new tab)questioning your own timelines

PG also cites Stripe in Schlep Blindness (opens in a new tab) (2012) as the idea thousands of programmers didn't see because payments looked painful.

Jason Fried and DHH (37signals)

Getting Real (2006, free online (opens in a new tab)) is the anti-VC counterpoint: small, profitable, opinionated software. Chapter titles and lines below are from the online text.

PrincipleWhat it meansChapterApply it by…
Build less"Do less than your competitors to beat them." Solve the simple problems; leave the hairy ones to othersBuild Less (opens in a new tab)scoping the MVP
Fix time and budget, flex scope"Never throw more time or money at a problem, just scale back the scope."Fix Time and Budget, Flex Scope (opens in a new tab)appetite
Fund yourselfoutside money brings constraints of its own; limited resources force you to launch soonerFund Yourself (opens in a new tab)asking what you can do with $20k instead of $100k
Pick the right customers"The customer is not always right." Basecamp focused on design firmsHire the Right Customers (opens in a new tab)a narrow first segment
Scale laterBasecamp ran on a single server for its first yearScale Later (opens in a new tab)what not to build
Half, not half-assed"build half a product that kicks ass"; Basecamp started with just messagesHalf, Not Half-Assed (opens in a new tab)cutting features, not quality
Start with noevery feature request gets "not now"; only requests that keep coming back get consideredStart With No (opens in a new tab)a high bar for new features
Forget feature requestsread them and throw them away; the important ones will keep coming backForget Feature Requests (opens in a new tab)feedback collection that tracks problems, not wishlists
Meetings are toxicthey break the day into incoherent pieces and convey little information per minuteMeetings Are Toxic (opens in a new tab)the same point as PG's maker's schedule (opens in a new tab)

Their later books extend the same stance: Rework (Crown Business, 2010) and It Doesn't Have to Be Crazy at Work (Harper Business, 2018), which argues for a "calm company" with sane hours and slow, profitable growth.

Tobi Lütke (Shopify): the trust battery

On The Knowledge Project (Ep. 41, "The Trust Battery" (opens in a new tab), 2018), Lütke described trust between colleagues as a battery rather than an on/off switch. Paraphrasing: when two people are hired into a company they start around 50% charged, because both passed the hiring process; each interaction charges or drains it; and the metaphor lets you talk about the state of trust without making it personal.

Apply it to co-founder and early-team relationships: autonomy follows charge. It's also a practical answer to the founder mode question of when to delegate.

Stewart Butterfield (Slack): sell horseback riding

Butterfield sent We Don't Sell Saddles Here (opens in a new tab) to the Tiny Speck team on 31 July 2013, two weeks before Slack's preview release, and published it in February 2014.

  • The hypothetical Acme Saddle Company can sell saddles on leather quality and price, or it can sell horseback riding, which grows the whole market and gives the saddles context.
  • Slack's version: "What we are selling is not the software product"; it is a result such as "a reduction in the cost of communication", and ultimately "organizational transformation". "We're selling better organizations, better teams."
  • The memo cites Andreessen's "only thing that matters" essay and says Slack was still before product/market fit, so the team must work "from both ends": improve the product and communicate its value better.

Apply it to launch copy and onboarding: describe the outcome, not the feature list.

Sara Blakely (Spanx): redefine failure

Blakely's account in Gillian Zoe Segal's Getting There: A Book of Mentors (Abrams, 2015), as excerpted in James Clear's 3-2-1 newsletter (opens in a new tab) (April 2024): her father would ask at dinner, "What did you guys fail at this week?", and was disappointed if there was nothing. She says it taught her to define failure as not trying, rather than as not getting the outcome.

Apply it to validation: a failed experiment that kills a bad idea cheaply is a win.

Jensen Huang (Nvidia)

Full sheet: Jensen Huang. In brief: go back to first principles and ask how you'd reinvent a thing with today's tools (Stanford GSB, 2024 (opens in a new tab)); chase "zero-billion-dollar markets"; run a flat company where decisions and feedback are shared in the open. His "ample doses of pain and suffering" line was said at the SIEPR Economic Summit at Stanford on 1 March 2024, not at the GSB.

Where they disagree

QuestionOne sideOther sideA workable reconciliation
How polished should v1 be?Hoffman: launch embarrassed; YC: launch nowSteve Jobs: insanely great; Chesky: design the perfect experience firstan embarrassing product is fine; an embarrassing experience isn't. Polish the core loop, cut everything else
CompetitionThiel: monopoly; competition is for losersPG and Friedman: err toward markets with competitorsa proven market, entered through a niche you can own
How fast to growHoffman: blitzscale once you have PMF37signals: calm, profitable, "scale later"; PG: stay default aliveblitzscale only with PMF and a winner-take-most market; otherwise grow on revenue
MoneyYC and a16z: raise to go fastGetting Real: fund yourselfdepends on whether speed decides the market
PlanningThiel: a bad plan beats no plan; Bezos: the long termAndreessen: the plan won't survive; 37signals: planning is guessinglong-term direction fixed, short-term plan disposable
DelegationPG and Chesky: founder modeHorowitz's peacetime CEO; conventional "hire good people and give them room"Lütke's trust battery: autonomy in proportion to earned trust
Customer requestsYC: talk to users constantly37signals: read feature requests and throw them awaylisten for problems, not for feature specs (see the Mom Test rules)
Decision speedBezos: decide at 70%Bezos too: one-way doors need slow careclassify reversibility first

Where each piece applies

Pipeline phaseFounder adviceWhere
Ideationregret minimization; schlep-heavy ideas (Stripe); sell the outcome (Slack)ideation
Validationfail cheaply (Blakely); many small bets (Bezos)validation
Product designPR/FAQ; 11-star exercise; half, not half-assed; fix time, flex scopeproduct design
Buildinglaunch embarrassed; scale later; six-page specbuilding the MVP
Launch & iterateCollison installation; handcrafted onboarding; sell horseback ridinglaunch & iterate
OperatingType 1/2 decisions; disagree and commit; trust batteryplaybook

Critiques and limits

  • Hindsight and survivorship. Every story here is told by a winner, years later. Cereal boxes, door-to-door recruiting and 3 a.m. support were also done by founders whose companies died; we don't hear their stories.
  • Scale mismatch. Bezos's letters describe a company with hundreds of thousands of employees. "Resist proxies" and "Type 1/Type 2" transfer well; "input metrics" and single-threaded leaders mostly matter later.
  • Context-specific success. 37signals' calm, self-funded model works for a profitable niche SaaS; Hoffman's blitzscaling works for winner-take-most networks. Each founder generalizes from the market they happened to be in.
  • Quotes drift. Famous lines get polished in retelling. Hoffman's aphorism is well documented by him; Huang's "pain and suffering" line circulates in several versions; Blakely's dinner-table question appears with small wording differences. Use the source, not the meme.
  • Founder mode vs trust. "Stay in the details" (Chesky via PG) and "delegate to trusted people" (Lütke, Horowitz) pull in different directions; neither has systematic evidence behind it yet.
  • Personal risk. Regret minimization and "fail at something every week" are easier with a safety net. Weigh them against your runway, dependants and the probability of outcomes.

Takeaways checklist

DECIDING
[ ] Is this a one-way or a two-way door?
[ ] Two-way: decide now at ~70% information
[ ] Disagreement: disagree and commit, or escalate
[ ] Big personal call: which choice will I regret at 80?
 
DESIGNING
[ ] PR/FAQ written before building
[ ] 11-star exercise on the core moment; pick the
    buildable star level
[ ] Scope cut to "half, not half-assed"
[ ] Time and budget fixed; scope flexes
 
LAUNCHING
[ ] v1 ships while still a bit embarrassing (not harmful)
[ ] Onboarding each early user personally
[ ] Copy sells the outcome (horseback riding), not
    the features (saddles)
[ ] Founders answer support
 
OPERATING
[ ] Six-page memo, not slides, for big decisions
[ ] Trust battery: autonomy matches earned trust
[ ] Meetings minimized; maker time protected
[ ] Failed experiments counted as progress

References