HooX
Music became my laboratory for exploring how far structured human-AI collaboration can go without losing the original intent.
It started with a simple idea: make some music for a gaming community. I could write the lyrics. AI could probably help with the rest.
That assumption turned into a much bigger experiment.
Today I approach a song almost like an architecture problem. Context moves between specialized tools and agents. Some steps need tight constraints, others need creative freedom. Outputs become inputs somewhere else. Decisions and learnings need to survive across iterations.
And every step raises the same question:
How much do I control — and where do I deliberately let AI surprise me?
Music makes that question unusually visible. A result can be technically good and still be wrong. The beat can work. The voice can work. The lyrics can make sense. And yet the song may completely miss what I wanted to express.
Conversely, AI can produce something I never imagined — and it can still carry exactly the intent I was looking for.
My job is not to specify every note. It is to preserve the core of the idea while creating enough space for capabilities I do not have myself.
That makes HooX a practical environment for exploring context engineering, agentic workflows, controlled variation, evaluation and creative collaboration.
The goal is not to make AI follow every instruction. The goal is to design an environment in which the right things stay constrained and the right things are allowed to emerge.
I learn something from almost every song.
And maybe this is where it gets really interesting.
There are points where I could spend months becoming better at production, mixing or mastering.
Or I could meet someone who already has that expertise — and perhaps their own AI environment — and see what happens when the two approaches meet.
Could one environment hand over not just files, but intent, context and decisions to another? Could a specialist take a strong prototype and turn it into something genuinely high-end without losing what made it work in the first place?
I don't know yet.
But I'd love to find out.
Because that would close another loop I care about: use AI and architecture to explore and prototype far beyond your own specialization — then let specialists take it somewhere you could not.
And perhaps, one day, we let the Gizmos talk to each other.
Somewhere along the way, this experiment also became an artist.