Keep on to blur preview images; turn off to show them clearly
any task that frontier AI can sort of do today, it’ll likely be able to do reliably one year from now
President & Co-Founder @OpenAI
any task that frontier AI can sort of do today, it’ll likely be able to do reliably one year from now
President & Co-Founder @OpenAI
查看全文：https://t.co/r55iuvSx6r
独立科技网站 - 蓝点网 / 感谢关注 订阅频道：https://t.co/xzeoUEoPcU 联系方式：https://t.co/LJK1g3biPp
因客观原因目前字节跳动和腾讯旗下的 AI 编程工具国际版均已下线 #Claude 系列模型，虽然引起部分开发者不满但也不可能恢复。删除 Claude 模型大概率是 Anthropic 的原因，9 月份该公司称将加强限制，对于中国企业或中国企业控股超过 50% 的企业，无论运营地点在哪都禁止使用 Anthropic 的服务。
查看全文：https://t.co/r55iuvSx6r
Here's my recent post about my new workflow and prompts for effectively using the agent mail system:
Former Quant Investor, now building @lumera (formerly called Pastel Network) | My Open Source Projects: https://t.co/9qbOCDlaqM
For my mcp agent mail project, I wanted to make a way to easily share the agent mailbox used by my coding agents for a real project so that people could really see in detail how it works, since screenshots only show so much. I've been working for a couple of days on a commercial "companion app" for the project and previously shared my workflow for that using 7 concurrent codex instances working as a team. After many hours of work, they've now exchanged over 1,000 agent messages, so I decided to use that as an extreme example. You can see the actual inbox for all the agents at the following link, which is a static site hosted on GitHub pages without using a traditional server: https://t.co/ZAbCMtFym0 It uses a WebAssembly version of SQLite and is pretty performant despite running completely in the browser. I also added a very easy-to-use command to the project so anyone can "export and share" their agent mailbox and automatically deploy it to GitHub using the gh utility (or Cloudflare pages). See the screenshots below for how that flow works. The "share" viewer matches the look and feel of the built-in python web UI and fully supports filtering, sorting, searching, and a nice "message thread" mode similar to the Gmail interface. And it's easy to update an existing mailbox share as your agents send more messages. It's useful on its own, but the real reason I added this feature was to show off the system when it's used in a real project, following the workflow I previously shared, where you use a detailed development plan and also have the agents use Steve Yegge's Beads project for task management. If you go through the actual sample messages at the link above, you can get a very good sense for how well the system works. The agents just figure things out and instantly know how to communicate effectively and self-organize to solve problems cooperatively. It takes just a couple minutes to set the system up on your machine using the one-line install command (see next tweet). Give it a shot and you will feel the AGI as your agents suddenly turn into a little development shop or startup team working for you.
Here's the one liner command to install it: curl -fsSL https://t.co/l05IgC4vbS | bash -s -- --yes What this does: Installs uv if missing and updates your PATH for this session Creates a Python 3.14 virtual environment and installs dependencies with uv Runs the auto-detect integration to wire up supported agent tools Starts the MCP HTTP server on port 8765 and prints a masked bearer token Creates helper scripts under scripts/ (including run_server_with_token.sh)