Summary
A local web UI (serve on localhost) that visualizes:
- Run history timeline
- Agent convergence heatmaps
- Diff viewer with side-by-side comparison
- Scoring breakdown charts
- Stats over time
This would be a major differentiator — no other AI coding tool provides visual analysis of ensemble results. Could be a simple HTML/JS page served via `thinktank dashboard`.
Why
Terminal output is great for individual runs but doesn't scale for analysis across many runs. A visual dashboard would make the scoring evaluation and convergence patterns much more intuitive.