AI Engineering Signal
Anthropic's Mythos cybersecurity model accessed by unauthorized users
Kimi K2.6 drops as a credible open-weight challenger to Claude Opus 4.7
A humanoid robot ran the half-marathon in Beijing in 50 minutes 26 seconds, beating the human record of 57 minutes 20 seconds
Claude Opus 4.7 ships worse than what it replaced
Claude Opus 4.7 ships 50% more expensive and immediately benchmarks worse than Opus 4.6 on long-context tasks
Microsoft's Copilot Studio prompt injection was patched
Anthropic's autonomous AI agents outperform human researchers on weak-to-strong supervision
Linux kernel finalizes AI-assisted code policy
Anthropic's Claude Mythos "thousands of zero-days" claim deflated
Claude Mythos Preview
Anthropic's Claude Mythos Preview
Meta Superintelligence Labs launches Muse Spark
This is the first model since GPT-2 that Anthropic has assessed as too dangerous for general release. The system card and red-team assessment are publ
Anthropic locked in multi-gigawatt TPU capacity via expanded Google/Broadcom partnership, while simultaneously disclosing a $30B run rate
Hallucinated citations are polluting scientific literature at scale
Subscribe