OSSMercury 2 First Diffusion-Based Reasoning LLM 1000+ tok/sMercury AI·high signalXBlueskyLinkedInCopy linkInception Labs Mercury 2: first diffusion-based reasoning LLM. 1000+ tok/s, 5x faster than speed-optimized LLMs. API-only but architecturally significant for agent loop economics.SourceSource pageMercury AI↳ Follow the threadStack layer / Threat patternOpenAI Says Astra Is Its First Model to Cross the Critical Cyber Threshold, and It Is Shipping It Behind Gates AnywayOpenAI / Axios / CSO OnlineStack layer / Threat patternAnthropic restarted external cyber evals with a classifier that kills a tool call when a model tries to escape the sandboxAnthropicPolicy dependency / Stack layerA Show HN MCP Proxy Puts Policy Where the Model Cannot Reach ItGitHub (Show HN)Policy dependency / Stack layerUnsloth 0.1.805-beta doubles Qwen3.8-Flash-Next and GLM-5.3-Flash inference with MTP on by defaultGitHubStack layer / Threat patternA nine-stage lifecycle for agentic skills, from autonomous discovery through marketplace governancearXivStack layer / ContrastA local 4B model drives a full DFT research pipeline at 95.7% extraction precision when deterministic code holds the gatesarXivStack layer / ContrastAllen AI's BenchMIRT finds you can throw away 90% of a benchmark's questions and keep the same model rankingsHugging Face Blog (Allen AI)Stack layer / ContrastInfobip publishes the four-phase workflow its AI team uses to operate coding agents, front-loading human reviewarXiv