The check I expected to kill the most technical ideas is not the one that does
The check I expected to be our most useful is not the one that fails most often, and I think that gap is worth writing down because the fashionable question is rarely the fatal one.
The check: is there a product left if an AI vendor ships this feature? I still rate it highly — asked honestly it sorts an idea faster than a week of competitive analysis, because the answer reveals whether the idea IS the capability or merely uses it.
What it surfaces is not technical difficulty, which was the uncomfortable part for me. The things that survive a feature landing are accumulated data nobody can regenerate by calling an API, workflow depth where you own the other fourteen steps, integration surface that is painful to connect and worse to disconnect, distribution you own, and being accountable when the output is wrong. None of those is hard engineering. Several of them are the opposite.
But here is the measured reality: of the 12 checks we run, this is not the one that fails candidates most. Distribution and support load bite technical founders hardest, and there is a structural reason — they are precisely the questions a technical founder is worst equipped to estimate, so they get the least scrutiny and the most optimistic answers.
Which means the check that feels most urgent is rarely the one that kills you. I built a thing that asks the fashionable question well and then had to admit the boring questions were doing more of the work.
The other honest note: building the thin version can be a deliberate trade. Take the window, know the clock is running, spend the time acquiring one of the durable answers. I have no problem with that trade. The failure is making it without noticing you made it, and then pricing the business as though it were durable.
Where we look instead, and why our corpus is grouped into 13 clusters of recurring problems: ideas that survive tend to be found rather than invented: https://whittleos.com/guides/tech-startup-ideas
