How to find billion-dollar startup ideas by being contrarian and right

Executive overview

As AI startup competition increases and obvious vertical opportunities get picked over, founders need a different approach to find good ideas. The two-year "gold rush" window that opened with AI is closing — every major vertical now has multiple competitors.

The path forward is finding contrarian bets: ideas that feel dangerous, scary, or unfundable, but are right because they start from first principles about what people actually need. Non-obvious isn't merely intellectually unclear — it feels uncomfortable.

The insight: nine out of ten people may call you stupid or crazy, but that one person who agrees is your signal you're onto something real.

The two-year window pattern

• Every major tech shift (internet, smartphone, AI) creates roughly a two-year gold rush of obvious ideas
• Uber, DoorDash, and Instacart weren't obvious during the iPhone launch era — no one predicted them
• Once the obvious ideas are picked over, founders must look deeper for a secret
• A secret isn't merely non-obvious — it feels risky, even dangerous to pursue

What makes an idea contrarian

• It has dead bodies stacked around it — previous attempts all failed
• It operates in a legal or regulatory gray area
• It targets a market that looks too small by conventional metrics
• It contradicts the current accepted playbook for how to build companies
• The existing "market" for it is wrong or hostile to the obvious approach

Case study: DoorDash

• Entered food delivery when Postmates, Seamless, Caviar, and others already existed
• The prevailing playbook at the time was the full-stack startup: don't just build an app, own the kitchen too (Sprig, Spoon Rocket)
• DoorDash's contrarian bet was doing the opposite — just the marketplace and delivery, no kitchen
• What looked like a crowded space with the wrong model turned out to be wide open

Case study: Coinbase

• Early Bitcoin culture was dominated by cypherpunks who wanted anonymous, anti-state payments (Silk Road)
• Brian Armstrong's contrarian bet was the opposite: partner with banks, comply with KYC/AML regulations, pursue mainstream users
• This made the product worse for the existing market and enraged the crypto community
• He was right that regular people would eventually want to trade crypto — the current market was not the real market

Case study: Uber and Lyft

• Zimride and Ridejoy were competing for long-haul ride-sharing via Craigslist-style matching
• Lyft (formerly Zimride) pivoted to short-haul daily rides using smartphones — a completely different use case
• The founders were genuinely worried they would go to jail when they launched Lyft
• Key insight: laws written before a major tech shift often don't reflect new reality and can be safely challenged when consumer benefit is clear
• San Francisco quality of life measurably improved once on-demand rides were available

Case study: Flock Safety

• Hardware product: solar-powered license plate cameras with edge computer vision
• Three strikes against it by VC logic: hardware, selling to neighborhood groups (small ACV), based in Atlanta
• TAM analysis capped the opportunity at ~$50–60M — a number that was simply wrong
• Garrett Langley ignored VC conventional wisdom and focused on what communities actually needed
• Growth came from an unexpected channel: local TV news coverage whenever a crime was solved
• Adjacent police departments would see the news and demand the product immediately
• Now valued at $7.5B and solves 10% of all reported crime in the US

Case study: OpenAI and SpaceX

• OpenAI launched to mostly negative press; the AI research establishment dismissed the team as unqualified
• Not publishing papers was treated as disqualifying — but papers were the wrong optimization target
• SpaceX was Elon Musk's fifth billionaire space venture — press assumed failure; reusable rockets were considered physically impossible by experts
• Both required founders to sustain conviction for years while the majority said they were wrong

Current contrarian bets worth examining

• Compound startups: Parker Conrad's Rippling model — building many interconnected products at once rather than a point solution. Hard in practice but AI makes it more executable. Campfire is doing this to displace NetSuite with a team of ~12
• Flipping the forward-deployed engineer: Palantir invented it, it became the default playbook — and default playbooks are ripe to flip. Gigamal has built an AI that does the FDE job in minutes instead of weeks
• CodeGen-driven enterprise sales: data migration that once took six months now takes weeks; time-to-value in enterprise can compress from a year to under a month
• Legal gray areas near recent tech shifts: open banking (data portability from banks), crypto regulation — laws written in a different era that don't reflect what users need

How to find your contrarian bet

• Start from first principles: what do users desperately want and need?
• Look for spaces with dead bodies — prior failures are signal, not stop signs
• Notice what feels uncomfortable or slightly illegal; great founders treat that discomfort as a signal
• Ignore conventional TAM math for nascent markets; the founders change what the market is
• Validate with users, not with VCs, Twitter, or TechCrunch — those signals reflect consensus, not truth
• You can't find the answer sitting at a computer; get out and talk to customers
• Work backwards from a growth goal to discover your real go-to-market, not the one that sounds reasonable up front