DAJAI · Sovereign AI
The Twin
DAJAI in machine form. Local. Sovereign. Runs on hardware you own.
What it is
The Twin is a 27B-parameter abliterated language model running locally on the Mac Studio M4 Max via MLX. 1M-token context window. No cloud round-trips. No vendor lock-in. It speaks for the artist when the artist isn't at the keyboard.
Where it lives
- BRAIN — Mac Studio M4 Max · 192 GB unified memory · MLX runtime.
- API —
/api/twin/chat· streaming · context-aware. - Surfaces — embedded chat widget across dajai.io and hellcatblondie.io.
Why it exists
Every dependency on a frontier API is a future tax. The Twin is the bet that an artist's catalog, voice, and decision pattern should live on hardware the artist owns — not behind someone else's rate limit.