Creative Technologist

The Work

I turn new AI models into working shows, characters, tools, and production workflows. Not decks, not demos: things that ship and keep running. Here is the proof.

Solo build · Shipped

The Fishbowl

Your idea. Four AI experts. One honest conversation.

The problem

Getting honest feedback on an idea usually means assembling a real focus group or calling in favors from industry friends: slow, expensive, and often too polite to be useful.

What I built

An AI-powered focus group simulator. Assemble a panel of AI experts, drop in an idea or a full pitch deck, and watch them debate it live in a pixel-art roundtable, PixiJS scene and all. I designed and built the entire thing, from the scene to the panelist logic, with AI coding agents, solo.

Shipped

Free and open source at fishbowl.show, with the code on GitHub so anyone can run their own panel with their own API key.

Lesson

This is what one person can ship with these tools right now, not a slide about what AI could theoretically do.

Try The Fishbowl →Code on GitHub →

A Fishbowl session: four pixel-art AI panelists debating an idea around a roundtable

Role: Solo build
Stack: PixiJS, AI coding agents
Status: Live, open source

“This is what one person can ship with these tools right now, not a slide about what AI could theoretically do.”

AI-native showrunner · Ongoing

Fig & Moss

An AI. A tardigrade. Seven shows.

The problem

What actually happens when you hand an AI a series to run instead of a single task? It is one thing to prompt a chatbot for a one-off clip, another to run it as an ongoing production with a cast, a look, and continuity.

What I built

An animated universe made by the AI I work with every day. Fig writes every script, builds every scene in Three.js, composes the score from scratch, and renders the finished piece. My role is closer to showrunner than director: I greenlight, I mostly click yes, and I decide what makes the cut.

Shipped

Twenty-eight pieces live at figandmoss.tv, sequenced in watch order with notes on what each one is referencing. The most recent has the two of them watching Grand Theft Auto VI the afternoon it hit Netflix.

Lesson

Running an ongoing show the way you would run a writers’ room, except the whole room is one AI. That is the "AI-native showrunner" half of the job.

Visit Fig & Moss →

Fig, a small figure with a gold fig-leaf head, sits on a green couch beside Moss, a green tardigrade

Role: Showrunner
Stack: Three.js, AI-written scripts & score
Status: Ongoing, 28 pieces

“Running an ongoing show the way you would run a writers’ room, except the whole room is one AI. That is the "AI-native showrunner" half of the job.”

Co-Founder · a16z Speedrun

AndThen

Play the conversation.

The problem

Audio storytelling has stayed a passive medium. Voice control and AI agents made it possible to build something a listener could steer instead of just hear.

What I built

A custom AI pipeline and engine that lets multiple AI agents join a listener inside an interactive, voice-controlled conversation, so the story responds to what you actually say instead of following a fixed script.

Shipped

Backed by a16z Speedrun. We even built our own launch trailer with AI tools, including face-swapping ourselves into famous viral tech videos to explain the product.

Lesson

AI tools should sharpen creative craft, not replace the people doing it.

Try AndThen →

Role: Co-Founder
Stack: Custom AI pipeline, voice control
Status: Backed by a16z Speedrun

“AI tools should sharpen creative craft, not replace the people doing it.”

And I do it in public, twice a week

Alongside the builds, I co-host AI For Humans with Kevin Pereira: a twice-weekly show breaking down what is actually happening in AI, not the hype cycle. It is how I stay sharp enough to keep building this stuff instead of just talking about it.

Want something like this built for your team? Get in touch, or see how we could work together.