Urgent.News

What's breaking now, across thousands of outlets.

More in AI

The Edit That Fixed 4 Tasks and Broke 1

AgentSelfEdit is an open-source sidecar that rewrites its own system prompt from execution feedback. It A/B tests edits and promotes only statistically-proven winners.

  • Four out of 26 tasks successfully fixed using the new prompt.
  • Six tasks still produced incorrect results due to ambiguous rule interpretation.

From Software Engineer to AI Engineer - Part 5: Scaling your tool belt

We learned about tools and wrote them ourselves. This is cute, but an application writing all of its own tools is not scalable.

  • Model Context Protocol (MCP) maintains catalog of AI tools
  • MCP server publishes catalog with tool names, parameters, descriptions
  • AI applications retrieve catalog via MCP client, process via model

More from Tuesday 1 September →