Credit: CaseBorne
She was so good in fact that she was soon promoted to commander, in another first.
，这一点在搜狗输入法2026中也有详细论述
This started with Addition Under Pressure, where I gave Claude Code and Codex the same prompt: train the smallest possible transformer that can do 10-digit addition with at least 99% accuracy. Claude Code came back with 6,080 parameters and Codex came back with 1,644. The community has since pushed this dramatically lower.
Add a cluster of points in one corner and watch that corner subdivide deeply while the rest of the space stays untouched. Then scatter a few points across the empty region and watch it split only where needed. The tree grows around the data.
Be the first to know!