Announced Oct 21 as a research preview for US Pro and Max subscribers, Duet runs on Claude Opus 5.5 in its own VM with a browser, checks your connected apps every few hours, sends a morning brief, and asks for approval before it sends, pays, signs, or deletes anything. It's a separate app on iOS, Android, and claude.ai, and its background checks count against your plan's usage limits.
Released Oct 21 on Hugging Face, Flint is a 40B mixture of experts model with 2.5B active parameters that Recursive says hit about 131 tokens/sec on an M4 Pro MacBook Pro and beat other sub 50B models on SWE Bench Verified (its own numbers). The company says its Eureka Machine research agent helped with nearly every step, and warns Flint needs big tasks broken into smaller steps…
Released Oct 21, new agent sessions now run on the Copilot SDK based harness, and a preview Compare Models mode in the Agents window runs one prompt across Copilot, Claude, and Codex harnesses with an optional judge model to score and synthesize results. Experimental on device MAI Code 1.1 Flash skips AI credits but needs an NVIDIA GPU with at least 96 GB of memory (Visual Studio Magazine)…
Announced Oct 21, GPT 6 Astra, GPT 6.1 Sol, and GPT 6 Luna are rolling out to select enterprise customers in Model Garden, billed through Google Cloud Marketplace against existing GCP commitments, and Codex CLI and Desktop can already point at Google's platform for inference (Google Cloud Blog). General availability is planned in the coming weeks, with Antigravity and Gemini Agent support to follow…
Released Oct 21, Moonshot's coding CLI now invokes plugin skills only by their plugin:skill name so same named skills stop shadowing each other, streams remote control traffic up to 100 MiB, and queues /compact while the agent is busy. Heads up: dangerous command guard = false now turns off every built in confirmation guard, including sensitive file and git control path checks (GitHub release)…
Generally available as of Oct 21, the Glean plugin's /glean run command writes an agent as a spec.yaml plus instructions.md in your repo, runs and evaluates it from the coding agent, then publishes it to Glean's managed runtime with its permissions across 275+ enterprise sources. Only Auto mode agents are supported; Workflow mode still needs Glean's visual builder (docs)…
Spotted something we missed? Start a thread.
