ByteBrief
We're a portrait publication through and through. Turn your phone back and your briefing picks up right where you left it.
(We tried widescreen once. It wasn't us.)
Researchers demonstrate a language model that rewrites its own weights using reasoning without gradients or training. An open 27B model improves 6.6 points on unseen problems and repairs a damaged copy to over 80% accuracy. This self-rewriting capability emerges only with scale, marking a step toward recursive self-improvement.
Tap to vote and see what everyone thinks.
Summary by ByteBrief