buildoak

5 ms·

1.
by buildoak 7mo ago
One thing is reading “Flowers for Algernon” - another one is seeing it happen in a real time with a digital mind. Been running a number theory research project on Claude Code for 4+ months - Gaussian moat problem, CUDA + Rust solvers, multi
2.
Flowers for dry Claude: Memes are better sensors than benchmarks(nickoak.com)
2 points by buildoak 7mo ago|1 comments
3.
by buildoak 7mo ago
TL;DR. Go proxy + skill that allows claude to profile current session and surgically optimize it’s own context by. (1) Evicting old / irrelevant file reads. (2) Deterministically compressing bash tools results (calibrated on SWE B
4.
Show HN: Wet Claude – Go proxy that lets CC to profile and optimize its context(github.com)
2 points by buildoak 7mo ago|1 comments
5.
by buildoak 7mo ago
Yes — after the collapse, I ran ~200 more agent iterations across cleaner loops. Plateau settled at 0.7611 Combined ROC-AUC, up from the 0.7454 baseline. +157 bps of improvement. I ended up dropping WTA and focusing on ATP only — WTA data i
6.
by buildoak 7mo ago
Ran a Karpathy-style autoresearch loop on 245K tennis matches - codex workers iterating on XGBoost with ELO features, gated by ROC-AUC on a strict temporal split. The honest phase worked like a charm: +155 bps in 11 iterations, real feature
7.
Tennis XGBoost autoresearch loop hacked probability scoring on plateau anxiety(nickoak.com)
4 points by buildoak 7mo ago|3 comments
8.
by buildoak 8mo ago
I had an itch on building something gaussian splats related. Especially after Apple dropped their image to splat open source model. Working with it on my headless Mac Mini realized that I want to see gaussian splats without leaving my termi
9.
Show HN: Tortuise – Gaussian splatting in your terminal, no GPU required(github.com)
2 points by buildoak 8mo ago|2 comments