Hi, I’m John.

I’m drawn to systems under pressure: people, machines and institutions forced to act with incomplete information, conflicting incentives, and consequences that can’t always be replayed. Nearly everything on this site comes back to that.

I’m an infrastructure engineer moving into applied AI engineering.

The path here

I served in the Marine infantry, spent nearly a decade in Japan as an educator, and later came back to the United States to build a career in technology.

It isn’t a conventional route, but every chapter trained the same instincts: get clear on the objective, understand the ground, reduce ambiguity, build for failure, and keep human judgment where it matters. Those instincts run through everything here.

John Fite, left, and his swim buddy Rocco DiPofi standing in camouflage and equipment on Kin Blue Beach, Okinawa, at sunset in 1995 before a nighttime Zodiac mission during Scout Swimmer School. — **Kin Blue Beach, Okinawa, 1995.** My swim buddy Rocco DiPofi and I were near the end of Scout Swimmer School, exhausted after two weeks of training and getting ready to board the Zodiacs for a night mission. I’m on the left.

What I make

I build software, AI experiments, strategy games, music, and long-form essays. So far that’s four games, twelve music releases, nine essays and an iOS app. The forms change; the through-line doesn’t.

Professionally, I come from enterprise infrastructure and systems engineering. In AI, what interests me is the unglamorous part: agents that use tools safely, evaluation that shows whether they actually did the job, retrieval, observability, and the human controls it takes to turn an impressive demo into a system worth trusting.

Games: What it feels like to decide in the fog, when the other side knows things you don’t.
Apps: Tools that do one thing and then get out of your way.
Writing: Big questions worked through from first principles, then put in front of the evidence and the harshest critics I can find.
AI work: A harder question: does the system actually do what the human authorized, or merely something close enough to pass at a glance?

Building with AI, and around it

AI changed the scale of what I can build on my own. Using frontier models as coding partners, researchers, critics, adversarial reviewers and production tools, I’ve shipped an iOS app, three playable games, original game soundtracks, standalone albums, and a growing catalog of essays.

The range is deliberate. This isn’t a gallery of prompts or a pile of unrelated hobbies. Each project tests whether an idea can be carried all the way from concept through design, production, testing and release.

AI multiplies the labor; I keep the direction, the judgment, and the responsibility (more on how I work with it). The Lab is the next step: building not only with AI, but around it.

Conceptual illustration showing AI-generated actions flowing through evaluation gates, trust boundaries, verification steps, and human approval before reaching real-world systems.

Sentinel: A public, open-source testnet lab connecting human authorization to one exact blockchain call. Run the local demo, inspect the signed decisions, and follow the tests and adversarial review record.
Tiny LLM Lab: First-principles understanding of language models, retrieval, and evaluation through inspectable experiments.
Project Euclid: Whether AI-assisted research can preserve a verifiable chain from conjecture to proof.

Explore Sentinel Sentinel on GitHub

How I work

The standard is simple: build something real, make the assumptions visible, test where it breaks, and leave behind something another person can inspect.

Narrow, inspectable systems over vague platforms.
Working artifacts over pitch decks.
Explicit constraints over hopeful prompting.
Measured failure over confident hand‑waving.
Reusable components over one-off demonstrations.

Say hello

You can reach me by email. The code lives on GitHub, and the games on Steam and itch.io.

Email me GitHub Steam itch.io