Agent swarms and the new model economics

Introducing Grok 4.5

Reward hacking is swamping model intelligence gains

Governing agent autonomy with Auto-review

What we’ve learned building cloud agents

Introducing Composer 2.5

Bootstrapping Composer with autoinstall

Continually improving our agent harness

Keeping the Cursor app stable

Better AI models enable more ambitious work

Speeding up GPU kernels by 38% with a multi-agent system

Better MoE model inference with warp decode