Simple: uv run script.py just works on a clean box/CI, the lockfile keeps runs reproducible, and my CI “install deps” step is way faster now.

Working on https://github.com/gobii-ai/gobii-platform open-source AI employees that live in browsers.

They’re always on. They log into real sites, click around, fill out forms, and adapt when pages change — no brittle scripts, no APIs needed. You can deploy one in minutes, host it yourself, and watch it do work like a human (but faster, cheaper, never tired).

Kind of like a “browser-use cloud,” except it’s yours — open, self-hostable, and way more capable.

Cool to see this coming out of FOSSA (ex FOSSA here :))

> Adding autocomplete to the prompt box gives me suspicion they somehow still do not understand best practices in using AI to write code. It is more crucial than ever to be clear in your mind what you want to do in a codebase, so that you can recognize when AI is deviating from that path. Giving the LLM more and earlier opportunities to create deviation is a terrible idea.

Agreed 100%

Any time there's LLM auto complete on the prompt (chatgpt has done this too!) I find it horribly distracting and it often makes me completely lose track of what I had in mind, especially on more difficult tasks.

garage

Add ZeroFS on top and get very low latency for frequently used data while bulk storage is remote S3.

The "stress" command is pretty good for this.

Glad we have an internal load balancing/failover LLM inference system, so this outage won't take prod down for us.

I can run GLM 4.5 Air and gpt-oss-120b both very reasonably. GPT OSS has particularly good latency.

I'm on a 128GB M4 macbook. This is "powerful" today, but it will be old news in a few years.

These models are just about getting as good as the frontier models.

> DNS (and networking in general) is a bit of a dark art.

Dynamic routing is fun :)

Time in the market beats timing the market.