NVIDIA open-sources SkillEvaluator; Hermes scans skill installs before confirm
NVIDIA open-sources SkillEvaluator after 300+ skill benchmarks; Hermes now advisory-scans hub installs for PII, secrets, and smuggling.
NVIDIA open-sources SkillEvaluator after 300+ skill benchmarks; Hermes now advisory-scans hub installs for PII, secrets, and smuggling.
Claude Code 2.1.236 adds ANTHROPIC_DEFAULT_MODEL, one-shot idle notify, tighter macOS sandbox denies, and auto-mode classifier improvements.
Grok 4.6 is GA on Amazon Bedrock: 500k context, xhigh reasoning, $2/$6 per 1M tokens, plus US and global cross-Region inference.
Anthropic’s Aug 19 notes make Agent Skills, the Files API, and Enterprise Admin user-management generally available without beta headers.
Cursor’s Aug 19 cloud-agent update adds event subscriptions, /goal, isolated subagent VMs, custom modes, and steering that waits for the next tool call.
Mastra documents AI SDK v7 helpers—chatRoute, workflowRoute, handle*Stream, and toAISdkStream—for TypeScript agent UIs.
Anthropic’s Python Agent SDK 0.2.140 adds MCP 2.x in-process servers and forward_subagent_text so nested transcripts can stream to the parent.
OpenAI’s Aug 18 Asana case study retells the Aug 7 Enzyme migration: four Codex agents, a five-sentence prompt, about $12K.
GitHub brought MCP allowlists, marketplace locks, managed OTEL, and Autopilot/Bypass locks to Copilot for JetBrains.
Claude Code 2.1.235 adds local prompt spellcheck, fixes LSP reconnect cache busts, and lightens background review/fix streams.