AI systems that survive contact with real work.
I design and build private operating systems for research, publishing, engineering, and decision-making: routing, memory, review gates, scheduled workflows, and human control.
Routing, memory, writing, and review gates moving through one working system.
My work sits between product design, automation, and applied AI. I care less about demos and more about systems that keep running when nobody is watching the terminal.
The throughline is operating discipline: clear inputs, useful memory, scoped model work, visible review paths, and proof that the workflow actually delivered.
Systems, not slideware.
These are the live surfaces behind the brand: the tools I use to route work, keep context, turn signal into writing, and ship code with review in the loop.
Prism
A Telegram-first command layer that routes messages into research, scheduling, memory, connectors, and hybrid Codex + Claude execution.
HQ
A private console for seeing which projects are healthy, which ones are stale, and what needs a human decision next.
The Take
A source-to-idea pipeline that turns curated video intake into notes, themes, packets, and reviewable publishing decisions.
ADW
A plain-English request becomes a plan, isolated branch, implementation pass, review trail, and pull request.
Notes from building past the demo.
Short essays on what works, what breaks, and what changes when AI systems have to operate inside real workflows.
Make AI useful after the demo.
I am open to systems work where reliability, taste, human control, and operational proof all matter.