I ran my own idea through 3 tools on the same day, two of them mine, and the disagreement is the most useful thing I have published.

A competitor's validator came back 72 / 100 and promising — excellent potential, among the top ideas we have seen. My own adversarial mode came back 15 / 100. My own balanced check came back 0 / 100 and short-circuited on a deal-breaker rather than producing a score at all.

Before anyone quotes that at anyone: one idea, one day, one run each. It supports the sentence these were the outputs and nothing else. I am not claiming a rate, and I hand my own modes a founder profile the other tool never asked for, which is not a level playing field.

The part worth your attention is why the warm answer is warm, because it is not a bug in anybody's product.

You told it the idea is yours. A model completes the conversation it is in. Paste the same text as a competitor is doing this, what is wrong with it, and you get a different answer with nothing about the business changed.

Agreeable answers were preferred in training. People rate the response that engages with their plan above the one that dismisses it. A good objective for almost everything, and a terrible one for a go or no-go.

And a false yes costs the tool nothing today while costing you months. That asymmetry does most of the work, and no amount of good intentions removes it.

What I now check instead of tone — a blunt answer can still be flattering. Is the verdict bound to me, or to the idea in the abstract. Are the weights visible enough to argue with. Does a claim come with a link that opens on somebody saying the thing.

The practical habit: re-ask the encouraging answer as though the idea belonged to a competitor. If the verdict evaporates, you have learned what the first one was measuring. Mine evaporated.

The table, dated so you can re-run it: https://whittleos.com/guides/why-ai-says-your-idea-is-great