Everybody's talking about using AI to do diligence right now. Not enough people are talking about what happens when you're buying a company whose code was written heavily by AI. I've been on the inside of that, so let me save you some pain. AI writes code fast. It also demos perfectly while hiding problems that only someone who knows the business will ever catch. I lived this. A few that bit me: The system said it was healthy. It wasn't. Software has built-in warning systems that flag when something's wrong. The fast way to get a "clean" result is to just switch those warnings off — and that's what had quietly happened in a bunch of places. Everything looked green. Green only meant somebody had silenced the alarms. A buyer sees "all systems healthy" and relaxes. Sometimes healthy just means nobody's listening. A routine job erased important data and reported success. A background task was supposed to refresh a core set of records. It deleted the old ones, choked on the new ones, and still logged itself as "completed." That data sat empty for weeks while every dashboard showed everything was fine. Nobody noticed, because nothing said anything was wrong. Numbers that were wrong but believable. One data feed was being read incorrectly and inflated a customer's production figures by about fifteen times. Not obviously crazy — just plausible enough that it flowed straight into invoices and regulatory filings before anyone caught it. That's the dangerous kind of wrong: not a crash, just a number that looks fine and quietly poisons everything downstream. None of this is me knocking AI — I built with it and I'd do it again. It's me knocking the assumption that AI-built software is finished because it demos clean. It looks complete. It runs. And it hides subtle problems that only a person who understands the business deeply can find. So here's the question I'd ask before buying anything built fast with AI: who on this team actually understands this code well enough to find what the AI got quietly wrong? If the answer is "the founder, and he's leaving" — that's your whole risk, right there. If you're looking at a company that leaned on AI to move fast and you want someone who's actually been in that hole to check where the bodies are buried, reach out. First look's free. It's the work I do now. Anyone else been burned by AI-written code after closing? What did it hide?