howen retweeted
Love building databases in Rust, especially for stream processing use cases? Apply for this role!
linkedin.com/jobs/view/44775…
howen retweeted
I might be the only person in the world that likes looking at Rust code. I think it's super pretty.
howen retweeted
i'll be back in sf from sept 29, figuring out what i wanna do next. if your company is hacking on something cool i'd love to chat!
moomin can offer coffee and fun little souvenir keychains for your time.
howen retweeted
We found some stuff. Actually, quite a lot of stuff.
Back in July, our team @rootxharsh, @S1r1u5_, and @iamnoooob managed to compromise multiple OpenAI employees’ ChatGPT accounts.
In theory, any ChatGPT connector can be affected. This includes Slack, github, email, and other services. 🧵
howen retweeted
I Built a mini LLM fully in Rust. No libraries other than ndarray + rand. Works just like the real thing, just a lot smaller.
I've got learnable embeddings, Self-Attention (not multi-head), Forward Pass, Layer-Norm, Logits etc..
Training set is tiny, but it can learn a few facts! Takes a few minutes to train fully in memory.
howen retweeted
The "whiteboard defense:" I should be able to pull you aside at any moment and ask you to explain any customer-facing system you've shipped. You should be able to clearly explain how it works and defend the decisions you made. This is my benchmark for responsible AI usage.
I don't expect line-level familiarity with the code. I don't care if you remember the exact function name or implementation detail. You may not even know it. I don't care.
But if I ask "why did you do X instead of Y?", "what happens if this actor behaves maliciously?", "what data structure did you use here and why?", or "where does this fail?" you should be able to answer confidently.
For PoCs, demos, experiments, whatever: I don't care. Generate 100% of it and understand none of it. Speed over quality every time in those specific scenarios.
But if you're shipping customer-facing work, you can't be shipping things you don't understand at a high level.
Balls are the perfect visualisation for benchmarks.
Inverted indexes are the perfect structure for text search.
"Inverted" here doesn't mean upside-down. Rather that it's an inversion of the original data. We take every distinct word from a set of documents and build a "posting list" of references back to every location where that word is used.
When structured right, this lets databases like Postgres do fast lookup of words and phrases.
TIN takes all this to the next level for Postgres. Span queries, fuzzy matching, wildcard matching, regex. All while strictly holding to Postgres durability, replication, and transactional semantics.
Also, it's FAST. Who doesn't love a good benchmark?
ahhh i miss cs3223
Query planning is hard.
Doing so in a massive, sharded database is even harder.
The blog we published today at PlanetScale was a labor of love. So many people from across the company helped with technical accuracy, writing, visuals, and making it the best it can be.
This was so much fun to be a part of. Go read. Get a little better at database query planners this evening.
