macOS container tool v1.0.0 released
34 points by dullmirror
34 points by dullmirror
Version 1.0.0 released. Release notes
"vibecoding" tag disclosure: Apart from a few commits "assisted" by LLMs[0], the .gitignore includes the Claude Code state directory[1] and the contribution policy encourages the use of "AI" tools[2].
[0] https://github.com/search?q=repo%3Aapple%2Fcontainer+claude+OR+copilot+OR+codex&type=commits
[1] https://github.com/apple/container/blob/6508acea81c193580fe07b9a61170caa50ba9443/.gitignore#L16
Tagging any software project that has an AI contribution policy as "vibecoding" is the worst possible conclusion to the ongoing debates around that tag's usage.
It's not just about the policy, it's about the culture. A lot of us don't want to see robot generated code and documentation and feel it makes Lobsters a less enjoyable website to use. Would you read blog comments sections if spambots suddenly started acting charming and funny, or would it get old quick?
At some point or another you have to make a distinction between “vibe coding” and using an LLM for auto complete or text transformations. The policy linked says:
You should be able to explain and justify every line of code or documentation that was generated or assisted by AI. Your submission should reflect your own understanding and intent.
That’s not vibecoding in my book. The policy says
We welcome thoughtful use of AI tools in your contributions to this repository.
That’s different from welcoming slop.
No, we don't have to let slop peddlers shift the overton window.
The overton window hasn't shifted, no one was melting down about spellcheck, autocorrect, or spam filters until very recently. You're the one trying to get everyone to think that autocomplete is suddenly bad.
one was melting down about spellcheck,
Possibly because no one ever thought that grep -f /dev/random /usr/share/words was a sane definition of "spell check" before
I share the feeling about vibecoded code but we should then put the tag everywhere nowadays. Do you know how much closed-source code by Apple is assisted by AI these days? I bet a lot. Should we put the tag on the next macOS release news?
Preferably yes, and then anything tagged vibecoding is automatically deleted. The amount of obvious and/or concealed AI-generated slop on the front page (even with the vibecoding tag automatically hidden) on any given day is astounding.
Honestly I would much rather read an interesting article that used a "spambot" for spellcheck or something than the 24/7 toplevel comments on every other post whining about "vibecoding". If the purpose of the "vibecoding" tag is so people who want nothing to do with AI can filter then (1) it seems like very few people are filtering considering every other post starts with a lengthy toplevel thread about vibecoding and (2) if you lower the bar for "vibecoding" to be anything however indirectly related to AI then you're going to filter out everything.
That’s an interesting question.
I think if I received value from them, and genuinely enjoyed reading the content, I probably would read them.
I’m not sure if it would get old, I suppose that all depends on quality, and the person behind these hypothetical bots.
robot generated code
please realize this combination of words makes you sound about one century over time
I mean.. the comment you replied to explicitly said the policy welcomes the slop.
You don't just stumble into that by accident. So either the project as a whole is pro-slop (in which case the tag is appropriate), or the maintainers didn't even bother reading their own policy (sure sounds like slop to me, in which case the tag is appropriate).
I haven’t seen anyone make the comparison yet, but this seems to be most similar to Lima.
Lima
Yeah: lima/colima. Similar to the point where I'm wondering if I should switch.
I've used this project in anger and can say with high certainly it's less mature than lima. I despise the gvisor sock shit in lima but otherwise it does what it says on the tin. This Apple tool is an MVP in my experience
I've been using it for a while, but I wish they'd pick more sensible defaults for CPU and RAM sizes. I first noticed things were slow, then that random tasks would crash, before realising that they were using under 5% of my RAM and less than half of my cores for containers by default.
Version 1.0.0 released
meta-edit: Hmm. I was prompted for a comment when submitting this entry because of a similar entry, but it kept complaining about "missing comment" when I clicked Preview, even though there was a comment (this one, without the edit), so I gave Submit a go. That worked.
Does anyone know why I would use this instead of podman?
It creates one VM per container, so you get better isolation (security and performance). Aside from that, they're pretty similar.