God thank you to this guy, this is the reality of AI. Those using it and building it daily know what the reality is. It’s not code, it’s a prediction engine, and 95% of that time the engine predicts right. The other 5% it’s fucking wrong. AI has been a great tool on top of our workflows as a double check, “Tell us in this PR if there are any security changes we should be aware of”. Works amazingly. Not critical to the process, adds a little label on the PR, we manually decide if we approve or not. But making it critical to systems? Horrible integrations. Suits don’t see that crap, it’s only us down here in the mines who are forced to see the realities.