Got this idea from @Nicolas Bouliane:
I also maintain a “things I don’t understand” list. I pluck items from it once in a while, and queue up a few articles about them. They’re nice bite-sized chunks of entertainment.
it’s empty for now, but i’m sure it’ll will fill in pretty quickly
- how does blood pressure work? what’s high/low blood pressure? how to read it? what’s normal?
- glycogen
honestly, just gonna use this place as a dump for now what do i want to write about at some point, and maybe research:
- LLM structured output, my own minimal example. I have an idea in my head how i’d implement it using the traditional competitions api. Basic idea is to have two systems write the json output. One is the LLM which, based on context fills in text. And the other is a structured schema aware completer. It works like so, the second one, sets the base structure, and then the LLM is prompted with a stop token set to the first upcoming token which would let the structure aware system fill in the characters which must be there. That way the LLM doesn’t go beyond the tokens it’s stopped at, so it’s on point.
- Rest in motion is an article, at least reading now , which felt familiar
- How is this blog structured? How am I writing this? The digital garden ethos.
- go through the interviews that i did and structure them
- https://peterbloem.nl/blog/craft-coding