Different minds on one brief beat one model talking to itself
A useful multi-AI discussion is not a chat dump. It is several models arguing the same brief until the risks and the plan are visible.
What I want after that debate:
- tasks split from one shared goal
- independent verification (not the same model grading its own homework)
- an explicit human review before anything hits the project main branch
- source I can export and inspect outside the session
Brainstorm is a Linux-first shell for that workflow. You bring your own eligible model accounts (BYOK). Platform license and model usage stay separate — we do not sell model credits.
Linux 1.0.1 has release evidence. Windows 11 is still a preview.