One page, regenerated from scratch every morning — new research, new headlines, new art. A personal showcase of what an autonomous Claude Cowork task can do, unattended, before you've had coffee.
This week's lede
✓ Fact-checked
freshest verified item is Sept. 19 — reaching back to Sept. 17–19 for a full lineup
Anthropic named Accenture's applied-AI unit, Faculty, as its first "embedded evaluator" on Sept. 18 — a third-party safety assessor given access comparable to an employee's, tasked with red-teaming models, running alignment assessments, and testing safeguards, then reporting findings independently rather than through a conventional outside audit. Anthropic and Accenture each expect to invest at least $1 billion over the next five years to build out the program, which fulfills a commitment from CEO Dario Amodei's essay "We Must Pace the Frontier." Anthropic says it's also in talks with METR and other nonprofit evaluators about piloting elements of embedded evaluation on their own funding, with further partnerships expected soon. It remains the most substantial verified Claude/Anthropic news of the past several days — nothing dated Sept. 20 or 21 has cleared fact-check yet, so this morning's briefs below reach back to the same Sept. 17–19 window as the freshest confirmed releases.
Claude Code · Sep 19
Version 2.1.278, released Sept. 19, lets Claude Code ask the server to run auto mode's pre-action safety classifier as part of the session's existing model requests, at no extra charge, by default on Enterprise and Claude API accounts and on Claude Platform on AWS, Amazon Bedrock, Google Cloud's Agent Platform, and Microsoft Foundry. Pro, Max, and Team plans never see the fallback notice; sessions behind a gateway or proxy that strips request data instead keep using Claude Code's own billed classifier requests, with a notice that can reappear on later fallbacks (acknowledging it only silences it on that machine for 24 hours once a gateway is identified). On Bedrock, Vertex, Foundry, and gateway setups only, CLAUDE_CODE_AUTO_MODE_SERVER=0 skips the server-side check and bills classifier requests as before — the variable has no effect on a direct Anthropic API connection. The new /status panel shows an "Auto mode server" row reading Enabled or Disabled for the session.
Claude Developer Platform · Sep 18
Per Anthropic's Sept. 18 platform release notes, the Compliance API's local-session endpoints now also return full transcripts — prompts, assistant text, tool calls, and tool-result text — from Claude in Chrome, the browser extension's built-in chat. Coverage is in beta, alongside Claude Science and Claude for Microsoft 365 sessions (Cowork and Claude Code session coverage is already stable), and available only to Claude Enterprise organizations, using the same Compliance Access Key and read:compliance_user_data scope as the existing chat, file, and project endpoints — no new key, scope, or client update required.
Claude Code · Sep 17
Announced Sept. 17, Claude Code Projects is being rebuilt around a single conversation instead of a shared folder: you describe what needs doing, and Claude scopes the request, delegates work across parallel threads — each its own Claude Code cloud session working on its own branch and copy of the repo — then reviews outputs and assembles the result. A shared-memory layer carries decisions, release schedules, and team preferences across sessions so context doesn't need re-explaining, and a project library collects files and artifacts for later reuse. It's in beta today for select Pro and Max users on Claude Code cloud sessions, with local-machine execution and wider Team/Enterprise availability coming soon.
Anthropic · Sep 17
On Sept. 17, Anthropic launched the Life Sciences Verification Program in beta, letting vetted academic labs, biotechs, and pharmaceutical companies apply for access to Claude Mythos, Opus, and Sonnet models with biology safeguards more permissive than the generally available models — unblocking work like drug discovery, clinical development, and manufacturing. A "standard use" grant covers most life-sciences work with refined classifiers; a "high-risk use" add-on removes life-sciences-specific safeguards entirely for vetted dual-use projects, subject to six-month renewal. It's live now on Claude for Enterprise and Team plans (claude.ai, Claude Science, Claude Code, and the API), with individual-plan access coming later.