Our AI kept rejecting big, crowded markets — and those were exactly where the winners showed up.

Our AI kept rejecting big, crowded markets — and those were exactly where the winners showed up. In our 2026-10-10 backtest of 100 real needs, of the 15 needs that later produced a product with 10,000+ users within 12 months, the judge had said "build" for only 2.

Why so conservative? Most of its "skip" verdicts were big, crowded categories — ad blockers, coupons, website themes, translation tools, crypto wallets. The reason was almost always "the incumbents are too strong, it's too crowded." Yet those are the needs where new products crossed 10,000 users.

A big market shouldn't be ruled out just because it's crowded. Crowding proves demand; what has to be new is your wedge, not the market.

Small sample, and this is an observation from a backtest, not proof — but it's a bias I've now removed from the pipeline: "too crowded" gets written down as a risk, not a rejection.

What's a crowded market you wrote off too early?