I approach AI as a producer first. In the same way a producer carries the schedule and the budget across years of high-stakes production, and answers for the room, I read the cost of a frame so I can protect the people who make it.
In my lab, I turn that producer's eye on the machine itself. I write the code and run the models myself, and the work goes into production.

Reads how an AI frame was lit, then hands you a recipe to carry that lighting into your own generations or relight the shot you already have. It even exports a Blender scene so you can rebuild the setup in CG.
Launch Capture the Light →Describe the shot. Get the model. Fifty frontier models rated on control, consistency, editing and where they break, each with its licence position attached. Then take that model’s prompt builder, paste it into any LLM, and get a finished prompt written in the structure that model actually responds to.
Open the Call Sheet →Runs your AI frames past a five-judge technical panel and scores them like real plates.
Convene the Jury →What does your voice look like? Ink & Voice paints the answer, pulling every colour and shape straight from the sound as it listens. Speak or play a song you love, and walk away with a video and a print that could only come from you.
Open Ink & Voice →The toolkit behind the build side. This is what I reach for, sorted by where it lives in the work.
I built this website along with the tools and research automation I use in my own work.
The instincts behind all of it took a career to earn. That story is here →