Lenny Rachitsky: And so when you talk about things getting stuck, it's like the AI agent just saying, I don't know what to do from this point or they introduce some kind of bug. Is that an example of getting stuck? Anton Osika: Yeah. It introduces some kind of bug and then it's not smart enough to figure out how to get out of that bug. Lenny Rachitsky: I see. And this is a common problem people have with tools like this is they get to a certain point and then it's like, "Well I don't know what to do. I'm not an engineer, here's a bug it's running into or the infrastructure's built the wrong way." And so it sounds like one of the paths to solving that is what you're describing is you make the AI smarter to avoid more and more of these places they get stuck. Another is people just learning how to get AI unstuck. This is something when we had Amjad on the podcast from Replit, he said that this is the main skill that he thinks people need to learn is how to unstuck AI when it runs into a problem. Just thoughts there, I don't know anything along those lines come up as I say that.
I'm not an engineer, here's a bug it's running into or the infrastructure's built the wrong way." And so it sounds like one of the paths to solving that is what you're describing is you make the AI smarter to avoid more and more of these places they get stuck.