I’d love to be able to set up agentic loops for my agents to build an app a day, improve them on their own and maximize token usage.
Sadly, no matter how much I try they always run into errors and...
I have a setup, which I am fine tuning, that rather solves stories, not apps at this level.
It can work end to end in a sense, that it has the whole loop of deving in it, with independent verification, review and fix.
But I am reviewing its plan before the implementation loop.
I did a simplification round yesterday. Now I have an orchestrator and 3 subagent (implementer, verifier, reviewer), and then the orchestrator just manages the loop.
It can work end to end in a sense, that it has the whole loop of deving in it, with independent verification, review and fix.
But I am reviewing its plan before the implementation loop.
I did a simplification round yesterday. Now I have an orchestrator and 3 subagent (implementer, verifier, reviewer), and then the orchestrator just manages the loop.
1