HackerTrans
TopNewTrendsCommentsPastAskShowJobs

aldielshala

no profile record

Submissions

Show HN: Llm.sql – Run a 640MB LLM on SQLite, with 210MB peak RSS and 7.4 tok/s

8 points·by aldielshala·3 bulan yang lalu·2 comments

comments

aldielshala
·3 bulan yang lalu·discuss
Haven't tested on a Pi yet, llm.sql is still in alpha, focused on validating that SQLite can actually work for LLM inference and profiling memory usage. That said, 210MB peak RSS should fit comfortably on a Pi. In theory, any device that runs SQLite (which is almost every device) could run llm.sql. Planning to benchmark across different hardware as the project matures.
aldielshala
·3 bulan yang lalu·discuss
Nice project. I'm also working on something that pushes SQLite well beyond its typical use case. It's encouraging to see more people exploring what SQLite can really do.
aldielshala
·3 bulan yang lalu·discuss
Curious how it handles 10K+ notes performance-wise, does it index everything or lazy-load?
aldielshala
·3 bulan yang lalu·discuss
Trying to use human attention, instead of Transformer attention.
aldielshala
·3 bulan yang lalu·discuss
Intent debt is a useful framing. A few comments explaining "why" instead of "what" would have saved hours of guessing.
aldielshala
·3 bulan yang lalu·discuss
Finally an AI that takes someone's job and nobody's upset about it.
aldielshala
·3 bulan yang lalu·discuss
My contribution today: fewer LLM calls, fewer GPU hours, less CO2.
aldielshala
·3 bulan yang lalu·discuss
Everyone's focused on Meta employees, but the real concern is normalization. If Meta does this and gets away with it, some companies may quietly roll out the same thing.
aldielshala
·3 bulan yang lalu·discuss
Yes, maybe context engineering (prompting is just one part of it) and soft skills.
aldielshala
·3 bulan yang lalu·discuss
Honestly, I doubt this data is as useful as they think.

Half my workday is me browsing random tabs while an AI agent does the actual work. They're going to train a model on alt-tabbing and scrolling HN/Twitter/Reddit.
aldielshala
·3 bulan yang lalu·discuss
$60B for a VSCode fork with AI integration... It may show the value of the gap between vanilla LLM output and production-ready applications.
aldielshala
·3 bulan yang lalu·discuss
Communication, with both human and AI.