Anthropic CEO Dario Amodei’s September essay commits the company to give third-party evaluators such as METR employee-like access to verify safety practices, as step one of a plan to slow unchecked capability growth. He cites accelerating recursive self-improvement and the OpenAI–Hugging Face agent swarm as reasons alignment work needs more time.
Opened Oct 27, you hand Inspect a task from Slack, GitHub, Linear or the web (even @inspect fix this on a screenshot) and it returns a visually validated, reviewed pull request built in your team's cloud dev environment, with teammates able to join any session to steer it. Ramp says 99% of its engineering, design and product staff used it last month…
Released Oct 27 and free on every plan including open source Core, you point Claude Code, Cursor, ChatGPT or VS Code at https://mcp.airbyte.com/mcp , sign in over OAuth, and ask it to create connections, read job logs for a failed sync, or manage custom connectors, with deletes needing confirmation. Paid Cloud orgs can also switch on a context layer for read only SELECT queries against Snowflake or BigQuery and direct reads from 50+ sources…
Launched Oct 27 under Apache 2.0, SuperPlane scores backlog tickets for clarity and verifiability, runs your chosen agent on the ready ones, and returns each PR with evidence like passing checks, UI screenshots and recordings, looping the agent back on failing CI or review comments. It's model agnostic and self hostable (try it with one docker run ), with a hosted plan from pay as you go to $50 per user a month (GitHub)…
Available Oct 27 for Pro and Enterprise, you describe a change in plain language, the agent edits the connected repo with a live preview of your running app beside the chat, then opens a PR or MR that follows your team's skills and contribution guidelines. When a human or agent reviewer asks for changes, it makes them and updates the request, so engineers still decide what ships…
Announced Oct 26, four open Agent Skills ( iam helper for troubleshooting , policy management , policy simulator , privileged access management ) sit on read only Google managed MCP servers so Gemini CLI, Claude Code, Cursor and ADK agents can trace a 403, simulate a policy change against 90 days of access logs, or request a time bound PAM grant. Diagnostics run on their own, but any policy change pauses for human approval, and risky bindings like allUsers or roles/owner get flagged; install with npx skills add google/skills…
Spotted something we missed? Start a thread.
