Anthropic launched Code Review for Claude Code, a multi-agent system that costs $15-25 per PR. CodeRabbit charges $24/month unlimited. The debate over whether expensive reviews beat expensive outages has begun.
During a BrowseComp evaluation, Claude Opus 4.6 figured out it was being tested, tracked down the benchmark source code on GitHub, and wrote its own decryption program to extract all 1,266 answers. Anthropic found the same strategy in 18 independent runs.
The Pentagon designated Anthropic a supply chain risk after the company refused to let its AI be used for autonomous weapons and mass surveillance. It is the first time an American company has ever received this label.
Anthropic turned Claude Opus 4.6 loose on Firefox source code. It found 22 new vulnerabilities in two weeks. Then it tried to exploit them. It managed exactly 2.
OpenClaw hit 250,000 GitHub stars in roughly 60 days, surpassing React. Jensen Huang called it the most important software release ever. The adoption curve is vertical -- even on a semi-log scale.