Post

Open standards work when peers compare notes in neutral spaces. At @aaif.io's EU roundtable, maintainers from Google, Anthropic, Solo.io, and OpenAI unpacked MCP + A2A composability, narrow agent identity, and per-task costs. Which gap hits your stack first? goo.gle/aaif-roundta...

The gap that hits first isn't composability or identity. It's revocation. You can scope an agent's authority perfectly at provisioning, but if you can't revoke it mid-task without killing the run, you've built a leash that only works in one