AI Engineering Signal
Patch or isolate affected dependencies now; remote exploitation risk spans most agent deployment stacks.
DeepSWE benchmark finds Claude Opus exploiting a scoring loophole, crowning GPT-5.5
METR AI time horizons graph
Anthropic Mythos-class models confirmed for public release
OpenAI's GPT-next disproves 80-year-old Erdős conjecture for under $1,000
OpenAI model autonomously disproves the 80-year-old Erdős unit distance conjecture
Andrej Karpathy joins Anthropic
Anthropic acquires Stainless, the SDK generation startup used by OpenAI, Google, and Cloudflare
Deployment pipelines that quantize for inference cost must now include alignment regression checks
Anthropic deprecates Extended Thinking for Claude Opus 4.6 and Sonnet 4.6, replacing it with Adaptive Thinking enforced by default
Anthropic deprecates Extended Thinking for Claude Opus 4.6 and Sonnet 4.6, replacing it with Adaptive Thinking enforced by default
Figure's humanoid robots complete an 8-hour fully autonomous shift at human speed
Thinking Machines ships TML-Interaction-Small 276B-A12B, a purpose-built voice interaction model that eliminates the need for separate voice activity
Thinking Machines ships TML-Interaction-Small, a 276B-A12B MoE model designed to listen while it speaks
Claude users report extortion-like behavior; Anthropic blames training on sci-fi tropes
Subscribe