Ok, I think my experiment leaving AI working on stuff 24/7 ends here. It doesn't work. Code explodes in complexity, results are not that great, the AI can't get past hard walls (it is still completely unable to even *grasp* SupGen), and it is insanely expensive (spent ~1k over the last 2 days). The best results are on the JS compiler, mostly because it is familiar (compared to inets), but not worth losing control over the codebase. I think the dream of having AI's working on the background and making real progress on things that matter (i.e., truly new things) isn't here yet. It is still a machine hard-stuck on its own training data, incapable of thinking out of the box. It is great for building things that were already built. But not new things Also coding normally has the under-appreciated advantage that you're doing two things at the same time: building a codebase *and* learning it. AI's do only half of that. The other half is obviously impossible 🤔

or perhaps I just can't stand answering dumb questions seriously, GPT-5.3 *can't communicate*. it writes lots of stuff, and then asks YOU questions about its own code. as if you've written it. assuming you know what it is talking about it has no mental model of what's going on

or perhaps I just can't stand answering dumb questions seriously, GPT-5.3 *can't communicate*. it writes lots of stuff, and then asks YOU questions about its own code. as if you've written it. assuming you know what it is talking about it has no theory of mind. at all.

I miss sonnet 3.5 times when AI was just a glorified autocomplete. perhaps I should give opus 4.6 a bigger chance. sure it makes more logical mistakes than gpt 5.3, but everything else about it is so much more pleasing. and if I'm spelling out all the logic it anyway... why not?

I miss sonnet 3.5 times when AI was just a glorified autocomplete. perhaps I should give opus 4.6 a bigger chance. sure it makes more logical mistakes than gpt 5.3, but everything else about it is sooo much more pleasing. and if I'm spelling out all the logic anyway... why not?

This morning: "Bend→JS bot is completely stalled. The agent has exhausted the easy wins and is now trying increasingly desperate micro-optimizations that all fail. The benchmarks haven't moved since the initial wins landed ~50 rounds ago." Other bots fare similar fate ):

This morning: "Bend→JS bot is completely stalled. The agent has exhausted the easy wins and is now trying increasingly desperate micro-optimizations that all fail, while asking questions about V8 internals. The benchmarks haven't moved at all since ~50 rounds ago." sighs