Platinum, per Ladybird's own sponsorship page, means $100,000+ per year. Money that, per the site, comes from "just people and companies who believe in an open web." Cloudflare, then.

Comments URL: https://news.ycombinator.com/item?id=46289143

Points: 3

# Comments: 0

Cringe. I suspect the same people who needed social comments and international media coverage to figure out that Christmas might actually be a nice time for some people are the ones who decided that video was appropriate in content and aesthetics. Also, that quote reads a bit like a machine desperately trying to understand humans.

Luckily, you can install a baby monitor quite easily on macOS by adding the following instructions to CLAUDE.md (or AGENT.md for opencode, etc.):

  Whenever you
  – make important progress
  – get stuck on the same issue repeatedly, or
  – finish the task
  use the `say` command to tell me what’s going on.

Comments URL: https://news.ycombinator.com/item?id=44809892

Points: 1

# Comments: 0

This question was triggered by this post: https://news.ycombinator.com/item?id=44196417. I’ve noticed that this topic comes up a lot in discussions lately.

Most of the projects I’ve seen so far seem to be more like proof-of-concept experiments—understandable, given the pace at which things are moving. As much as I’d like to adopt such tools, they often introduce dependencies on single-person projects, which feels too risky for my larger projects. Until things stabilize a bit, I’d rather set up a custom flow that gives me more control over the environment.

With that in mind, I’m curious how you approach this in your projects. Specifically:

- What approaches do you use to isolate AI agents? (e.g. container-based solutions like Docker/Kubernetes vs. cloud-based solutions)

- If you’re using the cloud: Are there good alternatives to GitHub Codespaces that offer less vendor lock-in but are still easy to manage?

- How do you balance a simple setup with the need for team collaboration?

- How do you handle large databases or datasets that AI agents might use or generate (e.g. storage, access, performance)?

I’d really appreciate hearing your insights, best practices, or any pitfalls you’ve encountered!

Thanks!

Comments URL: https://news.ycombinator.com/item?id=44198160

Points: 8

# Comments: 1

…

> Last week, the issue of the package uutils coreutils 0.0.30 successfully passed 507 tests (in the last issue of 506, in a year old – 476) from the standard GNU Coreutils test set.

If I’m reading this right then there are on average only one to five tests per utility (like cat, dd, …) which is a shockingly low number. Particularly when you consider how many options a single command alone might accept. And of course, this alone would not even remotely cover a critical part of all the circumstances that might vary.

I assume there is some kind of divide and conquer going on.