Dalton Caldwell and Michael Seibel from Y Combinator discuss why most AI startup ideas are derivative, how to escape consensus thinking, and why the best founders tend to be nonconformist people willing to work on things others have dismissed.
The Problem: Most Ideas Are Not Unique
With AI tools making it easier than ever to build, the bottleneck has shifted from execution to ideation. Yet most founders are not being encouraged to come up with unique ideas. Instead, they chase what recently got funded or what looks good to investors.
- Cloning whatever is trending on Twitter is low effort and unlikely to succeed
- Trying to "derisk" by copying a company that raised at a billion-dollar valuation provides no real signal — Dalton notes he often funded those companies for entirely different ideas that later pivoted
- Reading the tea leaves of YC batches is not a marketplace of pre-validated ideas
You're going to decide on what you're going to work on for maybe the next 10 to 20 years of your life by looking at a list of what other people are working on? Come on.
Look in the Discard Bin
Rather than searching for an idea nobody has ever thought of, founders should look for ideas that everyone has thought of and dismissed.
- Everyone applies similar filters to startup ideas, so truly novel undiscovered ideas are rare among billions of people
- The real opportunity lies in ideas discarded for reasons that may no longer apply
- Instacart is a classic example of a "discard bin" idea that succeeded
- Instead of seeking approval from friends, consider that universal pushback may actually be a positive signal — the Airbnb founders' parents were worried about them
Moving Away from Consensus
Unique ideas require moving away from the consensus of founders, investors, and friends. Several strategies can help:
Embrace Long Time Horizons
- Committing to something that takes 10+ years automatically filters out most competition
- Most people cut ideas that won't show results in two years, creating opportunity in longer-horizon projects
Study Non-Consensus Success Stories
- OpenAI started as a nonprofit research lab during the SaaS era — deeply weird at the time
- Anthropic had bizarre early funding from FTX — now seen as consensus-good, but was extremely non-consensus at inception
- Copying what worked for them in hindsight misses the point entirely
Be Wary of VC Content
- Before TechCrunch, founders had almost no VC content to consume — they thought from first principles
- Making life decisions based on podcasts is the definition of a non-unique idea
- The emphasis on "hundred billion dollar outcomes" causes founders to discard truly unique ideas that don't fit that framework
Solve Your Own Problem
The idea of being your own user has gone out of style, but remains powerful. Examples like Justin.tv/Twitch (camera on head) and Whatnot (marketplace for Funko Pops) would fail a TAM analysis but became massive successes.
Paul Graham built a Lisp website, got rich, and decided to do this really weird thing he made up from first principles. And as it turns out, that's the guy all these normie VCs look up to.
Weird People Have an Advantage
Nonconformist, quirky people naturally generate unique ideas because they think from first principles rather than seeking consensus.
- They don't need permission to ideate — their bigger challenge is filtering which idea to pursue
- People trained to calibrate consensus have the hardest time generating unique ideas — it has been "beaten out of them"
- As tech becomes more lucrative, it attracts more consensus-oriented people, potentially making the ecosystem less welcoming to the weird founders who drive real innovation
Practical Advice for Founders
- Brace yourself — working on something where many people think it's a bad idea is hard, but that is the experience of every founder who succeeded
- Be willing to stay committed for years, not months; investors can tell when someone is only 10% in
- If generating unique ideas is genuinely hard for you, consider joining a weird founder rather than forcing yourself to be the ideator
- Don't seek approval from authority figures — including Dalton and Michael themselves
If your filter is "do the people on these videos seem like they'd approve of your ideas," we're saying no. Don't be approval-seeking from authority figures if you want to actually have unique ideas.