Topic to App
Agents that find and evaluate app ideas.
Experiment
- Platform
- Agent system
- Kind
- Experiments
- Status
- Experiment
- Project year
- 2026
Product identity
The idea
Separate finding from judging.
The documented system separates scouts, a collector, a critic, a manager and a shared brief. The agents that collect complaints do not decide whether the idea is worth building.
The feedback loop
A rejection should teach the next search.
Verdict reasons feed back into the brief so the same rejected shape is not repeatedly proposed. Only the manager is permitted to reach the human.
Evidence boundary
A verdict count is not validation.
The August 2026 record contains 139 rejected candidates and an unrun control set. Those historical numbers do not establish critic quality or current operating status. This remains an experiment.