Anthropic put its AI coworker behind a US-only waitlist. The open-source clone… · @ReignSwords · TwKit

@ReignSwords

Anthropic put its AI coworker behind a US-only waitlist. The open-source clone is free, and it runs on your own machine. Open Duet: free, Apache 2.0, self-hosted. Named agents with their own memory, workspace, and routines. Any engine — Claude Code, Codex, Gemini CLI, even a local model via Ollama. Here's the first coworker setup I'd actually trust. Copy this: 📋 Name it + give it ONE job (example: morning brief, nothing else) 📋 Approval mode: Ask first. Always. 📋 One routine: runs on schedule, never approves its own actions 📋 One workspace folder it can touch. Everything else is off-limits. That's the whole trick. An agent that drafts all day and stops at the send button is a colleague. An agent that acts first and apologizes later is a liability. The honest limits: it's early, computer use is experimental and off by default, and you still pay for whatever model you plug in. No Gmail/Slack integrations yet either. My take: the waitlist isn't the story. The approval button is. Every lab is racing to build agents that act while you're away — the winner will be the one that knows when to stop and ask. Fellow AI builders — follow for more. Would you let an agent touch your inbox while you sleep, or is Ask-first the only mode you'd ever trust? 👇 🔵 Fellow Premium members, let's connect! I post about AI & tech — follow for more 👥

曝光 66 · 评论 7 · 点赞 1 · 书签 0