Ask HN: What are you working on (August 2024)?

I'm trying to figure out how to safely and effectively add hallucination-free LLM generation to medical care at my company: https://www.wyndly.com/

Been playing with RAG on my co-founder's medical expertise.

I have this creative writing website where writing friends can write branching fiction novels together. Sorta like choose-your-own-adventure, but more literary - third person past tense, actual characters and narrative arcs. Configurable number of choices at the bottom of every chapter, and as you're reading, if the choice label doesn't link to a chapter yet, you can write it. It's been floating around since 1996 or so, so my big project lately has been upgrading it. v1 was perl and gdmb files. v2 was php and codeigniter, and it's been limping along there for several years. When ChatGPT came out, I rewrote the whole thing for scala/play and that was super fun. Moved from ubuntu14 to ubuntu20, mysql5.5 to 8.0, etc. I'm done with the rewrite now but am still messing around with launching it, upgrading to ubuntu22 (done today), figuring out if I want to update jvm/scala/play before launching, etc. All this for only a group of six writers (so far)! But collectively we've written more than 425 chapters across a handful of stories, 450,000 words, and had a lot of laughs. I even wrote a snazzy graph visualization thing that animates the structure of the story maps. After I finish launching the thing on hardware/software that actually isn't EOL, I might open up membership to others.

I like the idea, but I have used Azure Pipelines, and the key differentiating it from GitLab.ci is its "service connection" that enables seamless integration with third-party services and a wide range of tasks (PowerShell, pscore, az, docker, etc).

I wish your dependency semantics yaml could be converted to Azure pipeline task, job, and stage similar to how typescript converts to javascript.

I am saying this because there are very complicated infrastructural problems that hundreds of people work on at big companies when creating a ci/cd I see you are not focusing on them yet but that DAG logic could be applied to existing azure pipeline yaml.

Reading chapter 3 of taocp (in vol 2). Randomness sparks something in my brain for some reason. It's nice to see Knuth already said something about what is a random sequence.

Trying to make my own thing that transcribes what I hear throughout my day. A lot of it is just finding an efficient stack (eg choosing silero instead of pyannote, azure vms instead of apis, whisper.cpp instead of faster whisper)