Engineering video
1:20 · 19 October 2026
Agents fail, try to fix their own mistakes, and can get stuck in loops contradicting themselves for up to a hundred turns. A hard limit stops the pipeline at this point to prevent wasting subscriptions and token limits.
Hear it at 0:00 →When agents get stuck, human engineers must step in to rethink code architecture, modularization, event-driven microservices, and missing context, as models still make fundamental mistakes.
Hear it at 0:42 →And the editing is whenever you run those tools and you run test, agents will fail and they try to fix their own stuff and they know fail again. And that's what we call turns. How many turns did an agent took to really fix this bug or feature? And at the beginning you think they do that three times or four times. But there are edge cases where no matter what model you use, you could use Fable, you could use Apple's 5.5, they drive themselves into loops, maybe because they miss some context, or they keep contradicting themselves, and I had agents trying to solve a certain bug a hundred times.
And that's my hard limit right now in the pipeline, which is 100 turns. If it takes more than 100 turns, stop and you go back to human review, because you're not it's going to burn through your whole subscription and token limit. This is where the real engineering comes in, because whenever you have one of those issues, I could spend a day just thinking about what happened and how do I change the factory or the code architecture? How do we modularize our code? Should we go more into even driven microservices?
Where is the context missing to really understand why couldn't the model fix the issue? So models they have a lot of knowledge, but you are definitely smarter as an engineer, okay? No matter what people say, model sites still make very dummy mistakes. And when they make mistakes, they go all the way to the bottom. So the picks are very high with AI, but the bottom is all of a very low, and that's why you are there.
And you can basically scale your judgment by embedding it into agents and skills once you figure out what the problem was.
