3 ms·

Here's what I find difficult to reconcile with my own experience. I've been using Codex in anger for the past 2 months or so (with gpt5.3-codex and then gpt5.4)

by pron 7mo ago

Here's what I find difficult to reconcile with my own experience. I've been using Codex in anger for the past 2 months or so (with gpt5.3-codex and then gpt5.4) on projects of different complexity. It is quite good at debugging, but the (non-trivial) code it produces is really bad. And I don't mean bad stylistically, but bad in the sense that Codex clearly won't be able to maintain it for long (which is how Anthropic's C compiler experiment failed) because it uses an approach of "success at all costs" where it always prefers fixes that treat the symptom rather than going back and rethinking the architecture as features are added. So the options are: 1. the program involved here is really trivial, 2. it hasn't evolved long enough for the agent to fail at evolution, or 3. others are not seeing what I'm seeing.

BobbyJo 7mo ago
This tracks with my own experience. I have to watch what the LLM does in response to any given ask, especially when adding new features to an existing system, because it will do things a different way every time until it no longer knows what patterns to follow.
sroerick 7mo ago
It is weird. I've had maintainable solutions on non-trivial code, but it does kind of require babysitting. Planning documents and detailed specs help. You get a feel for where the agent will want to take a shortcut and can devise ways to navigate around that. I also find Go works really well, and generally stays, if not exceptional, than at least maintainable. I've also enjoyed using OCaml, but I will say that I found the single worst function I've ever seen in a codebase in vibecoded OCAML. You might just try asking - "hey I'm having trouble keeping maintainable codebases - how can I structure this project in a way where the code will be stable long term". Sometimes getting the "software architect" role into the agent context is all it takes.
tintor 7mo ago
He used Claude Code, you are using Codex.

by pron 7mo ago

BobbyJo 7mo ago

This tracks with my own experience. I have to watch what the LLM does in response to any given ask, especially when adding new features to an existing system, because it will do things a different way every time until it no longer knows what patterns to follow.

sroerick 7mo ago

It is weird. I've had maintainable solutions on non-trivial code, but it does kind of require babysitting. Planning documents and detailed specs help. You get a feel for where the agent will want to take a shortcut and can devise ways to navigate around that. I also find Go works really well, and generally stays, if not exceptional, than at least maintainable. I've also enjoyed using OCaml, but I will say that I found the single worst function I've ever seen in a codebase in vibecoded OCAML. You might just try asking - "hey I'm having trouble keeping maintainable codebases - how can I structure this project in a way where the code will be stable long term". Sometimes getting the "software architect" role into the agent context is all it takes.

tintor 7mo ago

He used Claude Code, you are using Codex.