2025-02-18

Activity Log

hackers.town: 2025-02-18 Tue 09:49

Giving this song prominent placement in the playlist, times being what they are.

#NowPlaying

Float On by Modest Mouse

hackers.town: 2025-02-18 Tue 11:19

Making muscle memory mad today on the Model 100. Trying SPACE on the left, BACKSPACE on the right – again.

hackers.town: 2025-02-18 Tue 17:04

Your average project README is already scraping the bottom of the barrel when it comes to writing quality. No judgment about writing ability. People need editors, that’s all.

Seeing more LLM-generated READMEs, which take that barrel and throw it off a cliff.

(as opposed to LLM-reviewed, which could map to automated testing for clarity: better than nothing, worse than actual QA / editors; but not many folks looking at AI as a way to slow down and think things through)

hackers.town: 2025-02-18 Tue 18:35

TV captions describing the sound of my CPU fan.

hackers.town: 2025-02-18 Tue 19:46

She comes up just as I’m muttering under my breath at Ruff.

She: What’re you doing?

Me: Setting up my tool. I have a nice tool.

We:

hackers.town: 2025-02-18 Tue 20:18

Used Aider + one of those fancy DeepSeek-based 32B models (served locally via Ollama) to write some test code that I hadn’t gotten around to.

It took about five minutes before giving itself a timeout, and the results were very nearly syntactically correct. I have spent the last two hours fixing the output.

Huge productivity booster. If not for LLMs I would’ve had to write my own broken code.

I wouldn’t be surprised if the results were better with a hosted code-oriented LLM. All those options are paid and dubious – or “free” and dubious.

But I did not spend the last twenty nine years messing with busted Linux printer / audio / bluetooth / wifi / graphics / and-sometimes-keyboard drivers so that I could lease the ability to write code.

Also I’m broke. Same reason I got into FOSS in the first place. I can’t be paying for compute time or watching the drip-drip of free token allotment.