
Anthropic is making Claude meaningfully more autonomous with two additions to Claude Code and Claude Cowork.
First, Dispatch — which lets you hand Claude a task from your phone and maintain one continuous conversation across mobile and desktop. No context lost, no starting over.
Second, and more significant: computer use inside Claude Cowork and Claude Code. When Claude doesn't have the right connector for a task, it can now fall back to operating your actual computer — cursor, clicks, files, browser, developer tools — the way a human would.
Anthropic frames it carefully, as you'd expect. Connectors come first, explicit permission prompts are required, prompt-injection safeguards are built in, and sensitive applications are blocked by default. The safety scaffolding is visible by design.
More insights
All insights
Half the Market Is on Strike
September's IPO delays, record private rounds, and a secondary market where only 1 in 5 names has buyers and sellers. Our Q3 data, explained.
Read
Is the market immune to good news?
Nvidia posted its best quarter ever, yet AI sentiment barely moved. What August's LBX25 Index and private-market data say about where AI risk sits.
Read
If the model layer is commoditizing, why do the two frontier labs post the fastest revenue ramps in tech history?
July in private markets: open-weight models closed in, chips fell ~20%, yet Anthropic and OpenAI kept ramping. Here's where activity moved.
Read