well, its a multistage system

A signal generator is driven by touch sensors, generates a code. The code is hashed to soft prompt a uncensored offline AI, and the AI generates multiple outputs. they are inturn fed in a selector which also is driven by touch sensor. The selector picks one, which goes in a second uncensored MoE model, and amplifies the response.

All outputs are checked by a coherence governor.

Then the final output is ed to a final tool-operator stage. to drive TTS, or other things.