Codex Handwrites Weights for a 95.1% Digit Recognizer Without Training

A Reddit user had Codex handwrite neural network weights plus hand-coded rules—no optimizer or training—to build an MNIST digit recognizer that achieved 95.1% accuracy (1709/1797 test images).

2026-11-10 ~ 2026-11-10 · 2 related posts

• Codex builds a digit recognizer with no training, scoring 95.1% — GenJeppo · 2026-11-10
• Codex hand-writes neural net weights to build a digit recognizer hitting 95.1% without any optimizer — GenJeppo · 2026-11-10