devincrane

3 ms·

1. 1.▲
   by devincrane 7mo ago

   For context, this came out of building a small tool called runshift to experiment with the idea of an "agent control plane." The core idea is that agents can operate freely most of the time, but certain actions trigger a trust gat
2. 2.▲
   Running 9 AI agents and realizing we needed a control plane(runshift.ai)

   2 points by devincrane 7mo ago|2 comments
3. 3.▲
   by devincrane 7mo ago

   Running 9 agents right now. Five doing outreach, filtering, and content. Four helping with dev and scoping dev. The problem is you can’t realistically review everything they do. But letting them run completely unchecked means you only find
4. 4.▲
   by devincrane 7mo ago

   I’m building this right now, in a little beyond the POC, but would love your feedback. I have the framework down.