Agents are now negotiating permissions, fixing code, and deploying changes together
jasonkneen · x · 2026-07-26
A developer showed an agent-to-agent workflow where one agent on a remote system talks to a local agensis-agent on his machine.
The screenshot shows the agent being asked to:
- fix a state-directory bug,
- design a self-reload / rollback flow,
- verify permissions and health checks before writing code.
The key point is that the agents are not just chatting: they are negotiating access, proposing implementation details, and then fixing and deploying changes to the product itself.
More from coding & agent
- Cursor could regain the lead by betting on flexible models and lower inference costs — salahuddin · 2026-07-26
- Jack Dorsey’s Buzz pitch centers on shared compute for open-model communities — MarvinTBaumann · 2026-07-26
- Channels adds LiveKit-powered huddles for voice collaboration and task handoff — jasonkneen · 2026-07-26
- AWS, Google, Azure and Cloudflare are all building agent sandboxes — krishnan · 2026-07-26
- Agent workflows need deadline propagation, not just timeouts — blaizedsouza · 2026-07-26
- MCP release candidate adds stateless scaling and enterprise auth for agents — davemccollough · 2026-07-26