@JM0x5C

Touching grass instead of messing with my ARM code generator.. https://t.co/K6Sbpb8Jqt

I don't feel like I understand the codex workflow well. With Claude, I plan the thing, revise the plan, then execute. The plans are non-ambiguous and I can figure out code issues right away. People are telling me, with gpt, to just let the clanker work and not plan so heavily? But then it just keeps putting out slop. Is the idea to just let it rewrite it all constantly until some fixed point? My problem with this is that Claude is just writing code I would've written myself. I actually read the code, it's just doing the minutiae of typing it for me and fixing small bugs. With GPT I'm just supposed to accept slop until it de-slops itself? I just feel like fixing the issues GPT is creating is much worse than just writing the code myself. What am I missing? Claude's modus operandi seems like exactly how I write software. It's essentially a super google. GPT feels like having to actually "vibe" and I don't feel ok with that. I'm not a vibe coder, I read every line that goes in.

Claude when writing plans: I will perform A and B and C exactly in these ways Codex: I will do the thing. Why does codex generate such shit plans if you don't tell it to generate it with excruciating detail?

I'm not sure how now is much better. I did think speech used to be too policed back then, but now every other poster just calls me a beaner and I get racist comments like every week. Like how do we reach a middle ground here?

Another classic example of a suburbanite in the GTA pontificating about how the sprawl is totally fine. I'm from here, what do you plan to say now, that I'm pontificating because I don't like it being only dead space and single family homes mixed in with strip malls?