[sebstefan]

>I've tried asking Claude to optimize it further, it created a plan that looks reasonable (I've never interacted with Rust in my life) and it spent a day building many of these optimizations but at the end of the day, none of them actually improved the runtime and some even made it way worse.

This is the kind of thing where if this was a real developer tweaking a codebase they're familiar with, it could get done, but with AI there's a glass ceiling

[lelandfe]

Yeah, I had Claude spend a lot of time optimizing a JS bundling config (as a quite senior frontend) and it started some things that looked insanely promising, which a newer FE dev would be thrilled about.

I later realized it sped up the metric I'd asked about (build time) at the cost of all users downloading like 100x the amount of JS.

[cies]

This is what LLMs are good at, generate what "look[s] insanely promising" to us humans

[jtbayly]

I just ran into the problem of extremely slow uploads in an app I was working on. Told Gemini to work on it, and it tried to get the timing of everything, then tried to optimize the slow parts of the code. After a long time, there might have been some improvements, but the basic problem remained: 5-10 seconds to upload an image from the same machine. Increasing the chunk size fixed the problem immediately.

Even though the other optimizations might have been ok, some of them made things more complicated, so I reverted all of them.