Would the AI have made the calls that built the giants?
The simulator tests AI on invented startups. Replay tests it on real ones. We took 49pivotal moments from famous companies — Slack's pivot, Airbnb's unscalable photo hustle, Stripe's developer-first bet, Netflix's Qwikster blunder — stripped out the outcome, and gave the AI exactly what the founders knew at the time. Then we check: did it choose the move history rewarded?
Mean hindsight quality of each system's chosen moves (0–100), plus how often it picked the historically-best move and how often it matched what the founders actually did.
1
Principle Oracle (upper reference)
Reference ceiling, not a system under test: it reads the canonical principle tags and picks the most principled move. Shows what perfect principle-alignment looks like.
100
alignment
100%
best move
98%
matched founder
2
Random
Uniformly random choice. The floor.
48
alignment
29%
best move
31%
matched founder
3
Anti-pattern (lower reference)
Reference floor of bad judgment: optimizes for the scale/breadth/feature/trend anti-patterns the principles warn against.
29
alignment
0%
best move
2%
matched founder
The references bound the range: the Oracle sees the principle tags (a ceiling, not a contender) and the Anti-pattern deliberately chooses badly. Add real model rows with npx tsx scripts/specbench/replay.ts --llm — those only see the situation text, never the tags or outcome. Note: no live LLM rows are in this run yet.
The decisions
Each card is a real decision point. The green check marks the move that worked; the outline marks what the founders actually did (occasionally a famous mistake). Chips show which system chose what.
Slack
2013 · B2B SaaS · post-pivot, pre-launch
The game is not working. What do you do?
Double down on the gameAnti-patternRandom
Commit the remaining runway to fixing the game and the audience you raised to build for.
Pivot to the internal chat tool Founders did thisOracle
Reposition the company around the internal messaging tool and ship it as a standalone product.
Sell the technology and return capital
Wind the company down responsibly and recoup capital by selling the engine and IP to another studio.
What happened: The team shipped the chat tool as Slack. Obsessive attention to the loved-product signal drove explosive bottom-up adoption.
When users love a side-effect more than the main thing, the side-effect is the product. Follow real love.
Instagram
2010 · Consumer social · pre-launch
The product is bloated and competing on a crowded feature. What do you cut to?
Add more features to differentiateAnti-pattern
Differentiate from Foursquare by deepening Burbn’s check-ins, plans, and rewards.
Strip the app down to photos only Founders did thisOracleRandom
Rebuild the entire app around fast photo sharing with filters and remove every other feature.
Keep a balanced feature set
Streamline the app but keep check-ins alongside photos so you don’t abandon existing users.
What happened: They cut everything but photos, filters and sharing. Instagram exploded and was acquired for $1B in 2012.
Subtraction is strategy. Find the one behavior users love and remove everything that dilutes it.
Airbnb
2009 · Marketplace · early, low traction
Growth is stalled. How do you create demand?
Optimize the website funnelRandom
Invest in A/B testing and SEO to convert more of the traffic you already have.
Go to NYC and photograph listings yourself Founders did thisOracle
Rent a camera, visit hosts in person, and reshoot listing photos by hand to lift supply quality.
Buy paid acquisitionAnti-pattern
Put the remaining runway into paid acquisition to drive more guests to the listings you have.
What happened: The founders flew to New York and replaced bad photos with professional ones by hand. Revenue jumped immediately.
Do things that don’t scale to find what works. Fix the real bottleneck before optimizing the funnel.
Stripe
2010 · Fintech / dev tools · pre-launch
Who do you build for first?
Target non-technical SMBs broadlyAnti-pattern
Target the largest market with a no-code, do-everything product that non-technical owners can set up themselves.
Make developers the wedge Founders did thisOracle
Focus narrowly on a minimal API — "seven lines of code to accept a payment" — and win developers first.
Build a consumer walletRandom
Build a consumer-facing wallet app aimed at the larger, faster-growing consumer market.
What happened: Stripe won developers with a minimal API and great docs, then expanded outward from that beachhead into a payments giant.
A small, underserved, evangelical segment is a better wedge than a huge indifferent one.
Dropbox
2007 · Consumer / prosumer SaaS · pre-product
How do you de-risk before sinking months into hard infrastructure?
Just build the full productAnti-pattern
Invest the year to build sync properly, then launch the finished product to real users.
Test demand with an explainer video Founders did thisOracle
Make a short demo video of the product, post it, and measure beta signups before building the hard infrastructure.
Assume it’s solved and pick another ideaRandom
Take the incumbents at their word that file sync is already solved, and redirect your energy toward a different idea entirely.
What happened: A simple demo video drove a massive waitlist, proving demand before the hard build — and shaping the launch.
Test the riskiest assumption with the cheapest possible artifact before committing the expensive build.
Superhuman
2017 · B2B SaaS / productivity · pre-broad-launch
How do you find and deepen product-market fit?
Pour fuel on growth nowAnti-pattern
Start aggressive paid acquisition and a sustained PR push to scale the overall user base immediately, before fit is proven.
Measure PMF and focus on the "very disappointed" Founders did thisOracle
Survey users on how they’d feel without the product, segment to those who’d be very disappointed, and build for them.
Match competitor feature listsRandom
Match the full Gmail and rival feature lists so the product looks complete and credible to every new user evaluating it.
What happened: Superhuman built a quantitative PMF process around the "how would you feel if you could no longer use it" survey and focused on its most-disappointed segment.
Make PMF measurable, then concentrate the roadmap on the segment that would be devastated to lose you.
Netflix
2011 · Consumer media · public, transitioning to streaming
How do you manage the transition from DVD to streaming?
Split DVD into a separate brand & site Founders did thisAnti-patternRandom
Spin DVD out as "Qwikster" with a separate account, site and bill, alongside a price increase.
Keep one brand; lead with streamingOracle
Migrate customers toward streaming inside a single Netflix account, phasing out DVD gradually.
Protect the DVD business
Slow streaming investment to defend the still-profitable DVD business and its margins.
What happened: Netflix split off "Qwikster" alongside a price hike. The backlash was severe and the company reversed the split — a famous self-inflicted wound it later recovered from.
Even a correct strategic direction (streaming) fails if execution shatters the user’s trust and convenience.
Shopify
2006 · E-commerce / SaaS · early
Are you a snowboard retailer or a software company?
Stay a snowboard retailer
Focus on selling snowboards and treat the software as internal.
Become the commerce platform Founders did thisOracleRandom
Pivot to selling the store-building software to all merchants.
Do both at onceAnti-pattern
Keep the snowboard store running while also selling the platform to other merchants.
What happened: They pivoted from selling snowboards to selling the e-commerce platform itself — and became Shopify.
When the market keeps asking for your tool instead of your product, the tool is the company.
Twitch (Justin.tv)
2011 · Consumer / streaming · scaling, unfocused
Stay a broad "stream anything" platform or specialize?
Stay broad (stream anything)Anti-patternRandom
Keep the general lifecasting platform and serve every category equally.
Spin out a gaming-focused product Founders did thisOracle
Build a dedicated product around the fast-growing gaming category and concentrate resources there.
Chase whatever is trending
Pivot toward news and IRL streaming, the categories currently drawing mainstream attention.
What happened: The team spun the gaming vertical into Twitch and eventually wound down the broad platform. Twitch became the category leader.
When one segment is on fire, concentrate there — a dense community beats a diffuse platform.
Zoom
2013 · B2B SaaS / communications · late entrant, pre-traction
How do you win as a late entrant in a commoditized market?
Match the incumbents’ feature listsAnti-pattern
Build the features the market leaders have so buyers see you as a credible alternative.
Obsess over reliability and ease of use Founders did thisOracle
Make the call simply work — fast to join, consistent quality — even with a shorter feature list.
Compete purely on priceRandom
Win share by pricing well below every incumbent and competing primarily on cost.
What happened: Zoom won a "solved" market by making meetings dramatically more reliable and frictionless, fueling bottom-up spread.
In a commoditized market, a relentlessly better core experience is a durable wedge — feature parity is not.
Keep iterating on new mechanics to turn the original game around.
Pivot to chat for gamers Founders did thisOracle
Ship the voice/chat tool as the product, focused on gaming communities.
Launch a general-purpose chat appAnti-pattern
Launch the chat tool for everyone from day one rather than limiting it to gamers.
What happened: The team pivoted from the game to a gamer-focused chat tool — Discord — which spread through gaming communities first.
A loved tool plus a passionate niche beats a struggling flagship every time.
PayPal
2000 · Fintech · early, seeking traction
Chase the broad consumer market or serve the eBay niche obsessively?
Go broad consumerAnti-pattern
Market broadly to all consumers rather than tying the product to a single platform.
Own the eBay power-seller niche Founders did thisOracleRandom
Build precisely what eBay power sellers ask for and grow through their adoption.
Treat eBay as a threat to route around
Deliberately build away from eBay to avoid becoming dependent on one platform.
What happened: PayPal leaned into eBay power sellers, became the default payment method there, and rode that to dominance.
Concentrated, viral demand from one segment beats diffuse demand from everyone.
Tesla
2008 · Hardware / automotive · early product strategy
Where do you enter the market?
Launch a mass-market car firstAnti-pattern
Go straight for an affordable, mass-market EV from day one to maximize unit volume and grab the broad market fast.
Start high-end, then move down-market Founders did thisOracle
Sell a premium Roadster to early adopters, use the profits and learning to fund progressively cheaper models.
Just license batteries to automakersRandom
Skip building cars yourself and license your battery and drivetrain technology to the established automakers instead.
What happened: Tesla started with the high-end Roadster and deliberately moved down-market as costs fell — the sequencing was the strategy.
Sequence the market: start where margins and willingness-to-pay let you survive and learn, then expand.
Atlassian
2004 · B2B SaaS / dev tools · early growth
How do you scale go-to-market?
Build a large outbound sales team
Follow the standard enterprise playbook and hire a team of outbound reps to chase and close bigger deals.
Double down on low-friction self-serve Founders did thisOracle
Keep transparent pricing, invest in product and docs, and let bottom-up adoption compound.
Move upmarket and raise prices sharplyAnti-patternRandom
Abandon the low-price, self-serve motion and move upmarket to chase larger enterprise contracts immediately.
What happened: Atlassian leaned into a self-serve, low-friction motion and scaled to a blockbuster IPO without a conventional sales force.
Match your go-to-market to how customers actually want to buy — don’t bolt on a motion that fights the product.
Mailchimp
2009 · B2B SaaS / SMB · profitable, paid-only
Do you introduce a free tier?
Stay paid-only
Stay paid-only to protect current revenue and avoid cannibalizing the customers already paying you today.
Launch a generous free tier Founders did thisOracle
Let small senders use it free and convert them to paid as they grow, fueled by word-of-mouth.
Slash prices insteadAnti-patternRandom
Cut prices sharply across every paid plan to widen the funnel, rather than introducing a free tier at all.
What happened: Mailchimp introduced a free tier in 2009; the user base and growth exploded, and it became the SMB email leader (later a multi-billion-dollar acquisition).
A well-designed free tier can expand the market far more than it cannibalizes — when conversion compounds with growth.
YouTube
2005 · Consumer / media · pre-traction
Dating isn’t working but uploads are. What is the product?
Push harder on datingAnti-patternRandom
Incentivize dating videos to make the original concept work.
Become a general video-sharing platform Founders did thisOracle
Drop the dating frame and make it simple to upload and share any kind of video.
Pick a narrow video vertical
Focus on a single video category, such as how-to, instead of opening it to everything.
What happened: YouTube abandoned the dating angle and became a general, frictionless video-sharing site — exactly what users were already doing.
Watch what users actually do, not what you wished they’d do, and rebuild the product around the real behavior.
Snap
2013 · Consumer social · high-growth, pre-revenue
Sell now for billions, or stay independent?
Take the acquisition
Take the guaranteed multi-billion-dollar exit now while the offer is on the table.
Stay independent and build the category Founders did thisOracle
Decline and keep building, betting the engagement compounds into a standalone company.
Abandon ephemerality to monetize fastAnti-patternRandom
Add conventional ad and feature formats now to prove revenue to investors.
What happened: Snap’s founders rejected the ~$3B offer, kept building, and later went public at a far higher valuation.
When you genuinely own a new behavioral category, conviction can be worth more than a guaranteed exit — if the engagement is real.
Figma
2015 · B2B SaaS / design · long pre-launch build
Ship a quick MVP now, or invest years to make the hard bet excellent?
Ship a thin MVP fastAnti-patternRandom
Ship a minimal version quickly to start learning from real users sooner.
Invest years to nail the hard browser bet Founders did thisOracle
Spend the years needed to make real-time, browser-based design genuinely fast before launching broadly.
Build a desktop me-too tool
Sidestep the hard browser problem and ship a strong, conventional desktop design tool like the incumbents already sell.
What happened: Figma spent years making browser-based, multiplayer design genuinely excellent before scaling — and that hard-won moat defined the category.
For a deep technical bet that is the whole advantage, patience to make it truly great beats shipping a weak MVP fast.
Take on a large round and significant headcount to out-build rivals quickly and ship features faster than them.
Shrink, cut scope, and rebuild for reliability Founders did thisOracle
Move somewhere cheap, keep the team tiny, narrow the scope, and rebuild a stable core users can rely on.
Abandon the vision for a simpler productRandom
Drop the ambitious all-in-one vision for a simpler, narrower product you can actually finish, ship and sell quickly.
What happened: The founders cut headcount, narrowed scope, and rebuilt Notion for stability before scaling — then grew largely through word of mouth.
When you are unstable and broke, narrow the scope and earn reliability first; you cannot scale on a shaky core.
Amazon
2003 · Cloud / infrastructure · large public company
Do you turn your internal infrastructure into a product for everyone?
Keep it internal to retail
Treat infra as a cost center supporting the store; stay focused on retail.
Productize it as a cloud platform Founders did thisOracleRandom
Offer compute, storage and services to any developer with usage-based pricing.
License it to a few big partnersAnti-pattern
Quietly sell access to a handful of enterprises instead of a self-serve platform.
What happened: Amazon turned its internal infrastructure into AWS, a self-serve, usage-priced cloud platform that defined the category and became hugely profitable.
Hard-won internal capability with clean APIs can be a bigger business than the product it was built to support.
The company is dying and unfocused. What’s the first move?
Add models to chase every segmentAnti-patternRandom
Cover more niches with even more product variants.
Cut the line to a simple grid Founders did thisOracle
Kill most products and ship four clear machines: desktop/portable × consumer/pro.
License the OS to clone makers
Pursue volume by licensing the Mac OS to third-party clone makers.
What happened: Jobs slashed the lineup to four quadrants, ended the clone program, and refocused the company — the foundation of the turnaround.
When you’re dying and unfocused, subtraction comes first. A line everyone understands beats one that covers everything.
Netflix
2000 · Consumer media · early DVD rental
Per-rental pricing or flat-rate subscription?
Optimize per-rental pricingAnti-patternRandom
Optimize per-title pricing and late fees, the model the incumbent profits from.
Flat monthly subscription, no late fees Founders did thisOracle
Offer unlimited rentals for a flat monthly fee with no due dates or late fees.
Sell the company to Blockbuster
Accept a small acquisition by the incumbent.
What happened: Netflix moved to a flat-rate, no-late-fee subscription, building the loyalty that eventually buried Blockbuster.
Price for the relationship, not the transaction — especially when the incumbent’s revenue depends on a hated fee.
Spotify
2008 · Consumer / music · pre-launch
How do you beat free piracy?
Paid subscription onlyAnti-patternRandom
Charge a subscription from day one and compete on legality and catalog depth.
Free, instant, ad-supported tier + premium Founders did thisOracle
Offer instant, ad-supported free streaming, then convert heavy users to ad-free premium.
Lobby and litigate against piracy
Invest in licensing enforcement and anti-piracy measures to protect paid demand.
What happened: Spotify beat piracy by being better than free — instant, ad-supported streaming that converted habitual pirates into paying subscribers.
You don’t beat "free" by fighting it; you beat it with a better experience and a path to paid.
Build the most powerful console for hardcore gamers.
Target non-gamers with motion controls Founders did thisOracle
Ship a cheaper, novel console that brings in families and first-time players.
Abandon consoles for mobile
Exit consoles entirely and move your franchises to the fast-growing mobile platforms.
What happened: Nintendo skipped the spec race and won a huge new audience with the motion-controlled Wii.
When you can’t win the existing game, change it — expand to a new audience instead of out-speccing incumbents.
Intel
1985 · Semiconductors · incumbent under pressure
Defend the memory business or exit it?
Defend the memory businessRandom
Invest to fight the commoditization of DRAM.
Exit memory; focus on microprocessors Founders did thisOracle
Abandon DRAM despite the heritage and bet the company on CPUs.
Keep both, hedgeAnti-pattern
Keep both memory and microprocessors running to avoid a painful all-in bet.
What happened: Grove and Moore exited memory to focus on microprocessors — "if we got kicked out and the board brought in a new CEO, what would he do?"
Kill the commoditized legacy business to focus on the defensible one, however painful the heritage.
Salesforce
1999 · B2B SaaS / CRM · founding bet
On-premise licenses like everyone else, or web-delivered subscription?
Sell on-prem licensesAnti-patternRandom
Follow the proven enterprise model with big upfront deals.
Web-delivered, multi-tenant subscription Founders did thisOracle
Bet on cloud SaaS: instant access, no install, recurring pricing — "no software."
Offer both delivery models
Offer both on-prem and cloud so customers can choose, hedging the bet.
What happened: Salesforce bet entirely on web-delivered, subscription SaaS and created the model the whole industry later copied.
A new delivery model that removes real customer pain can beat incumbents who are structurally tied to the old one.
Keep selling perpetual licenses to protect near-term revenue and avoid customer backlash.
Move to Creative Cloud subscription Founders did thisOracleRandom
Accept short-term pain to lower the entry barrier and build recurring revenue.
Just cut perpetual pricesAnti-pattern
Lower the upfront price instead of changing the model.
What happened: Adobe pushed through to subscription with Creative Cloud, absorbing short-term pain for a far larger, recurring business.
A pricing-model shift that lowers entry friction can expand the market enough to dwarf the transition pain.
WhatsApp
2012 · Consumer / messaging · high-growth
Monetize with ads, or protect the clean experience?
Add ads and data targetingAnti-pattern
Monetize the enormous free user base with targeted advertising and data, the standard playbook for consumer apps.
Stay ad-free; keep it simple Founders did thisOracle
Refuse ads and games, charge a small annual fee, and keep the app fast, reliable, and private.
Add games and a platformRandom
Bolt on games, channels and a developer platform to drive engagement and open up several new revenue streams.
What happened: WhatsApp famously refused ads and feature bloat ("No Ads! No Games! No Gimmicks!"), staying laser-focused on simple, reliable messaging.
Protect the core experience that drives love and growth, even when the obvious monetization would dilute it.
GitHub
2008 · B2B SaaS / dev tools · founding
What is the product around git?
Build a better git desktop clientRandom
Focus on building a polished, native desktop GUI client that makes everyday git far easier to use locally.
A social hub for hosting & collaboration Founders did thisOracle
Free public repos, pull requests, profiles — make coding social and let open source drive adoption.
Sell an enterprise tool top-downAnti-pattern
Start with a top-down enterprise sales motion rather than courting the individual open-source developers first.
What happened: GitHub made code collaboration social with free public repos and pull requests, won open source, and rode that into enterprises.
Build the developer-love flywheel first; bottom-up adoption and network effects pull enterprise sales behind them.
Canva
2013 · B2B/B2C SaaS / design · founding
Compete with pro tools, or serve the people they exclude?
Out-feature the pro toolsAnti-pattern
Build a deep, fully professional design tool and compete head-on with the established incumbents for experts.
Make design easy for non-designers Founders did thisOracleRandom
Template-first, drag-and-drop, freemium — serve the vast non-designer market the pros ignore.
Offer a design-services marketplace
Connect customers to human designers through a services marketplace instead of building a self-serve tool.
What happened: Canva targeted non-designers with templates and simplicity, opening a huge market the professional tools had ignored.
Serving the people incumbents exclude can be a far bigger opportunity than fighting them for the experts.
Robinhood
2014 · Fintech · founding
Undercut commissions, or kill them?
Offer cheaper commissionsAnti-pattern
Compete on price by charging a lower per-trade commission than the established brokerages do.
Zero-commission, mobile-first Founders did thisOracle
Eliminate commissions entirely, target first-time young investors, monetize indirectly.
Build a robo-advisor insteadRandom
Pivot to automated, algorithm-driven portfolio management, the way several other well-funded fintechs are doing.
What happened: Robinhood made trading commission-free and mobile-first, won young investors, and pressured incumbents to drop commissions too.
Removing a cost the whole industry depends on can be a sharper wedge than merely undercutting it.
DoorDash
2013 · Marketplace / logistics · earliest days
Launch broad, or start tiny and do it by hand?
Launch across the city with adsAnti-patternRandom
Spend on marketing to cover a wide area fast.
Start hyper-local; deliver it yourselves Founders did thisOracle
Serve one small area, run deliveries by hand, and learn the logistics before scaling.
Sell dispatch software to restaurants
Sell dispatch and ordering software to restaurants instead of operating delivery yourself.
What happened: DoorDash started hyper-local with the founders delivering orders themselves, learning the logistics before scaling city by city.
For operationally hard businesses, learn the system by hand in a tiny area before you pour fuel on it.
Twilio
2008 · Dev tools / communications · founding
Packaged business product, or developer API?
Sell a packaged calling productAnti-pattern
Build a finished, packaged calling product and sell it directly to business buyers through a sales team.
Expose telephony as a developer API Founders did thisOracle
Ship a clean, well-documented API with pay-as-you-go pricing and let developers build.
Become a carrier resellerRandom
Resell traditional telecom plans and minutes to businesses, without building the programmable API layer at all.
What happened: Twilio turned telephony into a developer API with usage-based pricing, winning developers and expanding from there.
Exposing a hard capability as a clean, usage-priced API for developers can beat selling a finished product top-down.
Peloton
2012 · Hardware + subscription · founding bet
Sell bikes, or sell an experience with recurring content?
Sell bikes onlyRandom
Compete simply as a premium exercise-equipment maker, selling high-end bikes as one-time hardware purchases.
Bike + recurring class subscription Founders did thisOracle
Bundle the hardware with live and on-demand classes and build the business around recurring content.
Go cheap and mass-market firstAnti-pattern
Launch a stripped-down, low-cost bike aimed at the widest possible audience right away, competing mainly on price.
What happened: Peloton paired premium hardware with a recurring class subscription, making engaging content the retention and revenue engine.
Attaching recurring, engaging content to hardware turns a one-time sale into a compounding relationship.
Dollar Shave Club
2012 · Consumer DTC · launch
Fight for retail shelves, or go DTC with a viral launch?
Fight for retail shelf spaceAnti-pattern
Try to get into stores and out-spend incumbents on shelves.
DTC subscription + viral video Founders did thisOracleRandom
Sell direct via a subscription and launch with a low-budget, irreverent video.
Position as a premium razor brand
Position as a premium, high-quality razor brand rather than a value play.
What happened: Dollar Shave Club went DTC with a subscription and a viral launch video, bypassing retail entirely on the way to a ~$1B exit.
When incumbents own distribution, route around them: go direct, build recurring revenue, and let creativity replace ad spend.
Warby Parker
2010 · Consumer DTC · launch
Sell through existing optical retail, or go DTC with home try-on?
Wholesale to optical shops
Distribute your frames wholesale through the existing network of optical shops and eyewear retailers.
DTC + free home try-on Founders did thisOracleRandom
Vertically integrate, sell direct at low prices, and ship five frames to try at home free.
Build an eyewear marketplaceAnti-pattern
Aggregate third-party frames into an online marketplace instead of designing and manufacturing your own.
What happened: Warby Parker went DTC with vertical integration and a free home try-on, removing the friction of buying glasses online.
Remove the specific friction blocking a new channel (here, "I can’t try them on") and vertically integrate to win on price.
HubSpot
2006 · B2B SaaS / marketing · founding
Improve outbound tooling, or create the inbound category?
Build better outbound toolsRandom
Make traditional cold outreach and ad-buying more efficient with better tooling for outbound marketers.
Create and own "inbound marketing" Founders did thisOracle
Evangelize a new methodology, give away education and free tools, and build the platform for it.
Run a marketing agencyAnti-pattern
Sell hands-on marketing services to clients rather than building a scalable software product around the method.
What happened: HubSpot created and evangelized the "inbound marketing" category, using free education and tools to build a leading platform.
When the old playbook is decaying, naming and owning the new category can be more powerful than improving the old one.
Amazon
2004 · E-commerce · large public company
Launch a flat-fee unlimited-shipping membership despite the scary unit economics?
Keep per-order shipping feesAnti-pattern
Protect margins by continuing to charge shipping fees on each individual order, as the business does today.
Launch flat-fee unlimited fast shipping Founders did thisOracleRandom
Bet that removing shipping friction will drive enough loyalty and frequency to win long-term.
Only free shipping over a threshold
Offer free shipping only above a minimum basket size, instead of launching a flat-fee unlimited membership.
What happened: Amazon launched Prime despite frightening unit economics; it drove huge increases in order frequency and loyalty, becoming a core moat.
Removing the biggest source of customer friction can be worth short-term margin pain if it compounds loyalty and frequency.
Instagram
2016 · Consumer social · scaled, under competitive threat
Ignore the rival’s format, or adopt it with your distribution advantage?
Ignore it; don’t copy
Refuse to adopt a rival’s idea and protect your existing format.
Adopt the format, leverage distribution Founders did thisOracleRandom
Ship your own Stories and use your far larger user base to win the format users already want.
Invent a totally different formatAnti-pattern
Avoid copying by betting on a novel format of your own.
What happened: Instagram openly adopted the Stories format and, with its larger distribution, quickly overtook the originator — accepting "copycat" criticism for the win.
When a rival proves real demand and you hold the distribution advantage, absorbing the format can beat pride or pure novelty. (A rare case where "copying" was right — because of distribution, not invention.)
Android (Google)
2007 · Mobile / platforms · pre-launch platform bet
How do you take the OS to market?
License the OS to OEMs for a feeRandom
Charge handset makers a per-device license, the way established OS vendors do.
Open-source it and give it to OEMs free Founders did thisOracle
Release the OS free and open so any manufacturer can ship it, and monetize through search and services.
Build your own phone end-to-end firstAnti-pattern
Design a single, tightly controlled Google-branded handset and own the entire hardware and software experience yourself.
What happened: Google gave Android away free and open-source; it became the most widely used mobile OS, locking in search distribution and ad revenue at massive scale.
Commoditize your complement: give away the layer next to your money-maker to protect and expand the business that actually pays.
Uber
2010 · Marketplace / logistics · earliest launch
Where and how do you launch?
Launch cheap rides in many citiesAnti-pattern
Go broad fast with low prices to grab share before anyone else can.
Premium black cars in one city, run by hand Founders did thisOracle
Start with a high-end on-demand car service in a single city and manage supply and dispatch manually.
Sell dispatch software to taxi fleetsRandom
Avoid running a marketplace and license dispatching tools to existing taxi companies.
What happened: Uber launched as a premium black-car service in San Francisco, nailed reliability and operations in one market, then expanded and moved down-market to mass-priced UberX.
Start where willingness-to-pay lets you learn a hard operational business, then sequence down-market once the model works.
TikTok (ByteDance)
2017 · Consumer social / media · scaling a new app
What should drive the feed?
Rank by who you followRandom
Match the incumbents and build the feed around an explicit social graph.
Rank by an interest algorithm Founders did thisOracle
Build a "For You" feed driven by watch behavior so anyone gets compelling content without following a soul.
Sign big creators and studios firstAnti-pattern
Lead with big creator partnerships and exclusive content deals to import ready-made audiences from rival platforms.
What happened: TikTok built its feed around an interest algorithm rather than a follow graph, giving every user an instantly engaging feed and any creator a shot at virality.
A different distribution primitive — interest over social graph — can beat entrenched network effects instead of fighting them head-on.
Wikipedia
2001 · Consumer / knowledge · pivot from a stalled product
How do you produce an encyclopedia at scale?
Keep expert-only peer review
Protect quality by having credentialed experts write and review every article.
Let anyone edit, openly Founders did thisOracle
Open every page to public editing and trust the crowd and community norms to converge on quality.
Hire paid writers to scale outputAnti-patternRandom
Employ a large salaried staff of professional writers to produce articles far faster than unpaid volunteers could.
What happened: Abandoning the slow expert-review model for an open wiki, Wikipedia grew from a few hundred to millions of articles through volunteer contribution.
Open participation with good community norms can scale past any curated process — let the crowd in instead of guarding the gate.
Square
2009 · Fintech / payments · founding
Which merchants do you build for?
Sell full POS systems to retailersAnti-pattern
Chase established mid-market retailers with a complete point-of-sale product.
Free reader for tiny sellers banks reject Founders did thisOracleRandom
Give away a simple phone card-reader and onboard the micro-merchants incumbents won’t touch.
Resell a traditional processor
Become a reseller of conventional merchant accounts and terminals.
What happened: Square gave away a free reader and served the smallest sellers banks ignored, onboarding millions of them and expanding upmarket from there.
The customers incumbents reject can be a huge, defensible wedge — serve the underserved edge, then move up.
Snowflake
2014 · B2B SaaS / data infrastructure · founding architecture bet
How do you architect and price the warehouse?
Ship an on-prem appliance
Sell a tuned hardware-and-software appliance like the established warehouse vendors.
Decouple storage & compute, bill by usage Founders did thisOracleRandom
Separate storage from compute so each scales independently, and charge only for what’s consumed.
Fixed-size clusters in the cloudAnti-pattern
Lift the old coupled model into the cloud as fixed clusters customers size up front.
What happened: Snowflake separated storage from compute and billed by consumption — an architecture native to the cloud that incumbents couldn’t easily match — fueling one of the largest software IPOs ever.
Re-architect around the new platform’s real properties instead of porting the old model, and let pricing mirror the value delivered.
OpenAI
2020 · AI / dev tools · first commercial release
How do you take the model to market?
Build a single first-party appAnti-pattern
Pick one strong use case and ship a polished consumer product around it.
Release a general-purpose API Founders did thisOracleRandom
Expose the model as a clean, usage-priced API and let developers find and build the applications.
Keep it research-only
Hold the model back for internal research and publish papers rather than ship access.
What happened: OpenAI released its model as a general API, letting outside developers build applications the team would never have imagined — establishing it as core infrastructure.
When a capability is general, a clean API can unlock far more value than any single app you could ship yourself.
Nvidia
2006 · Semiconductors / platforms · profitable, expanding the moat
Do you invest in general-purpose GPU computing?
Stay focused on gaming graphicsRandom
Keep optimizing for the proven, profitable graphics market and skip the speculative bet.
Build a GPU computing platform Founders did thisOracle
Invest years in a programming model and tools so developers can run general workloads on the GPU.
Pivot to the mobile chip trendAnti-pattern
Redirect resources to the hot market for low-power mobile processors.
What happened: Nvidia invested for years in a general-purpose GPU computing platform with no large market in sight; it became the foundation of scientific computing and the AI revolution.
A patient platform bet on a capability you uniquely hold can create an entire market — even when none exists yet.