It can know if it was right or wrong the same way a dev does: by running tests, analyzing the failures, making adjustments, and rerunning until success. Or until it gives up. That won't stop it from going off the rails at times, so you also need a checkpoint system.

Can't argue with that.

I like approval voting. Nothing complicated, no ranking. Just "check all the boxes you would be okay with." Simple as that.

s/into/enough

Gemini has versions small into run on a phone too.

"I don't know what I'm doing, please suggest something that might or might not improve my company." Sounds like the entire industry is a perfect fit for AI.

Terminator: Smokey and the Bandit

Yeah but having the computer make adjustments fast enough to stay ahead of the instability (and failing) was a key plot point.

This is the plot of Voyager's "Timeless" as I recall.

You're using it wrong. If you are trying to vibe code at such a high level that ideas like "security" matter, you're going to have a bad day. The ideal case is when you have an existing codebase that requires a tweak. You already know just about what you want to do to it. Doing it yourself would take an hour or two. Using an AI might take a minute or two. But since you know exactly what you wanted, it's easy to verify the result. Plus, many of them are pretty good at "now add tests for that".

90% of the difficulty in using AI tools is knowing when to use them.

Google's AI search results are fully sourced.

It's time for lunch!

They cannot replace a human, but they can make a human more efficient.

Reliable and repeatable isn't the goal. The goal is to be able to verify a solution faster than one could create it....almost a P vs NP thing.