People moan about how badly made Claude code is but Codex seems much worse on every metric
It's like fighting fire by employing an arsonist.
It’s simple, deterministic stuff. At least most of the time.
I assume auto-updates are handled by deterministic code that's far away from anything agents can touch (self-modifying harnesses like Autolith notwithstanding). With the same behavior showing up in multiple high-profile harnesses (Codex, Claude Code), it feels likely there must be some actual reason for doing it this way.
It should be the harness responsibility to create / destroy things like worktrees and scratch spaces, and communicate to the agent (e.g. via system prompt) where the scratch space is and its purge schedule. Harness should also set appropriate env vars to make this work in tools.
> With the same behavior showing up in multiple high-profile harnesses (Codex, Claude Code), it feels likely there must be some actual reason for doing it this way.
I think fixing it is just not high on the priority list.
...are you sure it wouldn't consider your personal files garbage?
Although I just looked into it, and even Ubuntu 26.04 appear to default to tmpfs for /tmp storage, so I guess it's fairly universal now.
Not sure if I agree with that idea in principle... What's wrong with just wiping it intermittently??
But, I'm just wondering.
So you only snapshot the rest. (And since your source tree is under version control, finding stray garbage there is straightforward unless you checked it in)
Its the only way I trust to install anything from these companies.
Today, I generally spend a few hours on a spec, let the agent run for 4 to 6 hours and I'm 90% there, with maybe a couple of days of tweaks, then a couple of days of iterations coming from user feedback after that. I've stopped using skills almost entirely now, mainly using them just to keep discovery of MCP or API docs out of context until needed. I have a few bullet points on testing and style in AGENTS.md, but I'm pretty sure it would probably run pretty close to what I'm looking for even without those these days.
https://x.com/FlorianKluge/status/2041855675295318039
(Whether that's against the ToS is not clear.)
You can’t just use your Claude sub on a third party harness without getting blocked even if you use the Agent SDK. Apparently, Anthropic gives “credits” depending on your sub and you can use those _credits_ when you use third party harnesses or Claude -p mode. That usage is not the same Claude sub one would assume it is… It is all quite deceiving. Don’t get a Claude sub unless you’re okay with being locked into Claude Code…
E.g. OpenCode can be used with all three methods. Looks like it needs a plugin for the two Claude methods, though.
Anyone experiences with it?
it's minimal and somewhat opinionated, not integrated well with memory subsystems (although that's kind of the point)
my only real criticism is that last I played with it it deals with slow provider responses very poorly , so lots of locally hosted llms require a lot of tuning to get it to handle streaming/non-streaming/timers well.
way less minimal (but still pretty minimal) I prefer Pi; but different markets I think.
hax fit my openbsd laptop well, espescially since Pi and related derivatives are kind of un-spokenly incompatible with the BSDs.
Also one brief note for Codex users specifically. hax can work out-of-the-box as codex-cli drop-in replacement (since it can use the already authenticated codex token), so it's super easy to try it - basically it will be a zero-config experience.
So far I have very positive experiences, the results are generally better (and not just faster) than in other harnesses with the same models. I've done the same (user) prompts for tasks in a range of them, and compare the outcome to eval between them. Simple (harness) beats complicated one many times. I do wish it could use MCP though, but not as often as I think, and CLI tool usage often compensates on this front with little extra effort)
That is now my usuall go-to harness for things usually. Still figuring out the edges though (when we get off the happy path, like models reaching their max context, or such...)
It's very light, recommed you try it out, and see if it fits your toolbox.
And it worked. I'm switching to this unless i find a reason not to.
my main issues with maki are that the line editor (what you use to enter your prompts) is very barebones, copy-paste is broken (no middle-click paste, or copy to PRIMARY; implicit copying to CLIPBOARD that can only be used with a mouse), and scrollback is not native. So it's not a real cli, but some sort of weblike tui.
I tried 3code; it does not mess with selection, but code printed in the conversation has | markers in every line, and is thus uncopyable in practice. Don't know how to use that.
Also, it feels that chatgpt.com and aistudio.google.com are the buggiest websites I have seen for long time. Why it is always something related to LLMs?
But in the same fix, they also removed the reasoning-sample, which kinda sucked.
Chat sessions are not metered by token usage but rather a fairly high limit of messages per day or something.
The longer chats you have, the more context and token input. So I think that discourages longer chats, and encourages users to start new chats.