🧭 Claude Meetings Is Now Live on Google Meet — Joins Calls as a Visible Participant, Builds Docs and Whiteboards in Real Time
Anthropic today confirms the Google Meet integration for Claude Meetings is rolling out in beta to paid Team and Enterprise plans. Claude appears in the call as a named participant with its own panel, takes notes and transcripts, answers questions from attendees, and creates artifacts — documents, whiteboards, and slides — while the meeting is still in progress. No copy-paste or post-meeting summary workflow; Claude works concurrently with the participants.
How it works in practice
- Invite via add-to-call: From any active Google Meet, click Add people, search for
claude@meetings.claude.ai, and accept the join request. Claude's panel opens on the right of the meeting screen for all participants.
- Live artifact creation: Say "Claude, start a project plan for what we just discussed" and a draft document or whiteboard populates in Claude's panel in roughly 10–15 seconds. Participants can open the artifact as a full tab alongside the call.
- Notes and transcript: Claude generates a structured notes file at the end of the meeting. Notes and transcripts are free during beta on Team and Enterprise. Docs, whiteboards, and other artifacts created during the call are free through February 10, 2027.
- Limitations at launch: Not available for organizations with zero data retention (ZDR), HIPAA-ready configuration, or customer-managed encryption keys (CMEK). Google Meet is the only supported video platform today; Zoom, Teams, and others are on the roadmap.
Getting the most from Claude in a live call
Claude Meetings works best when participants direct it explicitly rather than waiting for it to infer tasks from ambient conversation. Brief, imperative prompts during the meeting — "Claude, summarise the decision we just made", "Claude, add a risk column to that table" — produce far tighter artifacts than asking for a summary at the end. For recurring standups, pin a Claude Meetings project so it carries forward context between sessions: prior blockers and action items surface automatically in the next call's notes.
Claude Meetings
Google Meet
Cowork
beta
Team
Enterprise
real-time
🧭 Claude Code 2.1.323 Ships Experimental Code Mode — Claude Now Writes JavaScript to Compose Its Own Tool Calls
Claude Code 2.1.323 lands today with an opt-in Code Mode that fundamentally changes how Claude sequences tool use. Instead of emitting tool calls one at a time and waiting for each result, Claude writes a small JavaScript program that invokes Claude Code's tools in a coordinated loop. For workflows with many interdependent tool calls — multi-file refactors, build-then-test cycles, iterative search-and-replace — this can collapse latency significantly by batching what would otherwise be sequential round-trips.
What's new in 2.1.323
- Experimental Code Mode (opt-in): Enable per-session via
/experimental code-mode on, or toggle it off for an entire organization in Organization Settings → Claude Code → Experimental. When active, Claude's tool calls arrive as a JavaScript snippet rather than individual JSON payloads. The snippet executes inside Claude Code's existing sandbox with the same permission model — no new permissions are granted.
- Plugin version range pinning: Plugins can now declare
engines["claude-code"] in plugin.json. A plugin outside its stated version range won't be installed, updated, or loaded. install and update automatically pick the newest compatible version. This prevents silent breakage when Claude Code ships a breaking plugin API change.
- Plugin TypeScript type-checking:
claude plugin validate now type-checks a plugin's TypeScript hooks module and reports each finding as a warning. Add --strict to fail the validation on any warning — useful in CI gates before publishing a plugin.
- Unknown model context window default raised: Claude Code now assumes a 1 M-token context window for model IDs it doesn't recognise, up from the previous 200 K default. Relevant for teams pointing Claude Code at private or fine-tuned model endpoints.
Code Mode — what to watch
Code Mode is marked experimental for a reason: how it interacts with --dangerously-skip-permissions and fine-grained permission rules is not yet fully documented. Anthropic recommends enabling it in a sandboxed dev environment before rolling it out to production pipelines. Expect iteration on the JavaScript API surface before it graduates to stable.
Quick test: does Code Mode help your workflow?
Open a Claude Code session on a known multi-step task (e.g., "rename all usages of getUser() to fetchUser() across the repo, run tests, and report failures"). Enable Code Mode with /experimental code-mode on, then run the same task. Compare the wall-clock time and the number of intermediate confirmation prompts. If your task involves more than ~6 sequential tool calls, Code Mode should show a measurable speedup.
Claude Code
v2.1.323
Code Mode
experimental
plugins
tool use
performance
🧭 Anthropic IPO Book Closes Multiple Times Oversubscribed — Final Pricing November 18, Nasdaq Trading Starts November 19
With five days until the Anthropic IPO pricing event, the order book for the 825-million-share offering at the $70–$78 price band is already reported to be multiple times oversubscribed. Seven strategic partners — including Google and Nvidia — have separately agreed to purchase $26 billion in non-voting stock at the IPO price, providing a floor under demand. At the top of the range, the offering would raise approximately $64.4 billion and value Anthropic at up to $1.41 trillion on a fully diluted basis.
Key dates and structure
- Final pricing: On or about November 18, 2026 — after market close.
- First day of Nasdaq trading: November 19, 2026, ticker ANT.
- Share composition: 801,351,354 shares from Anthropic (new issuance) + 23,648,646 shares from certain existing stockholders.
- Strategic anchor: The $26 B non-voting strategic tranche from Google, Nvidia, and five other partners is separate from the public float and not subject to the same lock-up schedule.
- Valuation range: $1.26–$1.41 trillion fully diluted at the $70–$78 band.
What "multiple times oversubscribed" means in practice
An oversubscribed book doesn't guarantee a first-day pop — allocations get cut, and institutional buyers who receive fewer shares than requested sometimes buy aggressively in the aftermarket, lifting the open price. It does, however, give underwriters (Morgan Stanley, Goldman Sachs, J.P. Morgan) confidence to price at the top of the range or marginally above. Retail investors using brokerage IPO access programs should expect allocations to be small or zero; the institutional demand compresses retail access in heavily oversubscribed offerings.
IPO
Nasdaq
ANT
roadshow
funding
valuation
Google
Nvidia