Lenny RachitskyHost of Lenny's Podcast | Author of Lenny's Newsletter

AI still needs humans

28 May 2026 · 1 min

Watch on YouTube

0:51

Lenny Rachitsky built an app called Proof using AI code generation, but it crashed repeatedly. When he tried to use OpenAI's Codex to auto-fix the bugs, it created cascading errors instead. He ended up sleep-deprived with bursitis, needing two senior engineers to fix what automation couldn't.

Chapters

Automation is a lie in the sense that every time you automate something, you need a human on top of it, like making sure that it's working well. We kind of miss that in the model discourse. And one of the reasons is benchmarks make it look like AI is more autonomous than it is. Because I've been feeling this paradox a little bit. I've been feeling we have so much automation and so much AI, and I also work way more. Have this app Proof. I just vibe-coded it on the side. When we launched it, because it was completely vibe-coded, it just started going down and I couldn't fix it. And it was very embarrassing. I had a lot of egg on my face. Just every 10 minutes, the Servers would go down and people were looking at me and I'd be like, I don't know what's going on. Like Codex, fix it. And Codex was like, I don't know what's going on. Or really Codex was like, uh, I do know what's going on. I fixed it. And then it would cause 4 other errors. And then you're just going around in a circle and I wasn't sleeping and I vibecoded so hard I got bursitis on my elbow. I got actually 2 different senior engineers to fix it independently.