Armin Ronacher ⇌

It is time for us to invite vibe-coders into our programming communities. lucumr.pocoo.org/2025/7/20/the-…

Another great example of Claude helping you figure stuff out. Yesterday I incorrectly assumed Omarchy did not have pam set up correctly because sudo did not work. Reinstall did fix it. But I just locked myself out. Claude confirmed it and told me how to change the lockout policy.

This year we’re renting a camper van during the summer with the kids. To improve on the camping experience (van or not) I got a proper battery pack for once :)

So today I have to say this again: UUIDv7 primary keys are a valid choice. They eliminated the performance problems of random UUIDv4 values. And if you need to shard one day, you can easily move rows to other shards because you won‘t have clashing numerical ids.

qwen-code is a fork of Gemini CLI. Because it legally can be. So let me plead again: please @AnthropicAI, open source Claude Code. This will be better for the ecosystem!

Another model release that advertises a claude-code specific auth system. All of this is so damn exciting :)

One of the things I learned from Omarchy is how just outdated my neovim/vim setup is. I really missed a few years of massive improvements in that editor.

Van acquired :)

Having the weirdest Linux issue though. For some reason on X in Chromium the backspace key does not support key repeats for me? Not sure what is going on.

Omarchy acquired and running properly. Definitely hitting some rough patches but quite impressed overall.

claude-code is an excellent onboarding body to modern linux.

Oh hello. @dhh made do it :)

Everything we _should_ have done (albeit low ROI) to make it easier to maintain software is about to have a lot more reason to invest into it. Docs, testing, isolation, etc. will finally become common place simply because it will enable us to achieve more via agents.

I fully support this. (Also purely selfishly I just want to normalize driving minivans. Need more choice in the market) x.com/annamlulis/sta…

Quite frankly the ability of Opus to root cause some more gnarly issues is great. Yes, I could have found this myself (dumb bug caused by a refactor), but still. And it used the available tools perfectly to validate the fix. When I see it do stuff like this it makes me happy.

It's a good first step that there is now a growing consideration in some communities that it's okay for newcomers to use AI, but using agentic coding for senior engineers is not okay yet. I think it is a much more mature response than to completely outright reject it.

Latest best feature of Agentic coding for me is how good they are at removing code. I feel a lot more confident now accepting sub-optimal code if it's in one piece of functionality, even if it's split across the codebase.

Relates well to my post: this is another good example why agentic coding can be really powerful and transformative. Use the LLM and write programs you can review. No need to worry about the LLM calculating wrong. And it involves (so-far) non programmers. x.com/badlogicgames/…