maxrumpf

6 ms·

1.
by maxrumpf 22d ago
The easiest way to strip complexity is to expose simple tools to an agent model like SID-1 that can use them well. It makes more of an effort for hard questions, and little effort for easy ones. (found of sid.ai so obv biased)
2.
by maxrumpf 1mo ago
SID-1 co-author here: was a lot of fun writing it. want to chat? just firstname at domain.
3.
by maxrumpf 2mo ago
> "compatible" instead of "comparable." It boggles my mind how you can train a frontier model but not write a tweet without an obvious typo.
4.
by maxrumpf 3mo ago
does it support fts/vector on edges of the graph?
5.
by maxrumpf 4mo ago
are you a maintainer on npm?
6.
Training SID-1 to beat GPT-5 at search with 1k+ QPS RL(turbopuffer.com)
12 points by maxrumpf 4mo ago|0 comments
7.
by maxrumpf 4mo ago
congratulations!
8.
by maxrumpf 6mo ago
We think this is a pretty sad day for research: Some context about Chroma's model. https://x.com/maxrumpf/status/2037365748973384154?s=20
9.
Scaling compute for retrieval by 5 OOMs: SID-1 tech report(sid.ai)
4 points by maxrumpf 9mo ago|1 comments
10.
by maxrumpf 9mo ago
This is the tech report for a model I helped work on. I'm biased, but it turned out very well. We essentially let the model learn to retrieve like a human would: Make a first search, read the results, and then make another. This lets t
11.
JIT *Coding* not just compiling(maxrumpf.com)
1 points by maxrumpf 2y ago|0 comments
12.
by maxrumpf 2y ago
This comment section is scary. Hacker news advocating FOR nanny technology?!
13.
by maxrumpf 2y ago
Excel is awful in almost every way, but I just wish more software was as customizable. I can get (even more) customization by using pandas etc., but it's usually much slower and you get much less of an intuition about the data.
14.
N=1 software and the unreasonable effectiveness of Excel(maxrumpf.com)
3 points by maxrumpf 2y ago|2 comments
15.
by maxrumpf 2y ago
I imagine color consistency will be such a pain here.
16.
by maxrumpf 2y ago
The weirdest thing people do is make up criteria that YC supposedly uses to reject people. There was such a huge diversity in our batch: From 20 y/o to 40+. Foreign, domestic. Credentialed, not credentialed. $1M rev run rate, $0 run ra
17.
by maxrumpf 2y ago
The abstract and the rest of the paper don't really match imo. It's not really allocating more to some sequences, but just introducing ~dropout. Might be different sides to the same coin, but was still a weird read.