About | James Dishman

The Short Version

I design things at the intersection of AI, creativity, and human decision-making. I am finishing an MS in User Experience Design at Kent State University. Before that, I got a BA from the University of Pittsburgh in English, Philosophy, and Political Science, which turns out to be a surprisingly good foundation for thinking about how people actually use things.

I produce electronic music. I founded the Pittsburgh Ableton User Group with Paul Miller. I serve as the direct liaison to Ableton's international community team.

What I Actually Believe

Technology should make humans better. I design the part where that actually happens.

That is not a tagline. It is a filter. Every project I take on, every system I build, gets measured against that. If the technology does not make the person using it more capable, more creative, or more connected to the thing they care about, I am not interested in building it.

I see design through Don Norman's frameworks as a standing lens. Affordances, signifiers, mapping, feedback, conceptual models. Not because they are trendy (they are decades old), but because they work. They are the vocabulary for asking whether something actually serves the person in front of it.

I am influenced by Hunter S. Thompson's radical honesty, not his substance abuse. The part where you look at the thing clearly and say what you see, even when it is uncomfortable. That matters more in design than people want to admit.

The Differentiator

I do my best work when I know the subject deeply. The Ableton evaluations in my portfolio are credible because I have lived inside that ecosystem for years. The Oblique Oracle exists because I cherish the I Ching, not because I saw a market opportunity. Clarence is not a demo project. It is a system I use every day, one that runs 25 cron jobs while I sleep and has over 1,500 memories in its knowledge database.

That depth is what I bring to a team. I learn the domain, I learn the users, and I build things that reflect both.

What is Next

I am in the capstone phase of my MS at Kent State. After that, I am applying to PhD programs in Human-Computer Interaction. The research questions I care about sit at the boundary of AI and creativity: what happens when the tool starts having opinions, and how do you design for that without pretending it is simple?

If you are working on research problems or building tools that need someone who understands both the design and the system underneath it, I would like to talk.

Building This Site

This site was built over about three days in late March 2026. The stack is Next.js, TypeScript, CSS Modules, deployed to GitHub Pages with Cloudflare in front of it. No templates. No themes. Built from scratch.

It was a collaboration between me and Clarence, my AI agent system running on Claude. I directed every decision. Clarence executed. That distinction matters. I chose the typography, the layout, the voice, the color palette, the information architecture. Clarence wrote the code, ran the builds, managed the deployments, and kept track of forty things at once while I was out in the field swapping meters.

How It Actually Worked

I would review the site on my iPad and phone between stops on my meter route, then send corrections via Telegram. Clarence would pick them up and push changes. Sometimes this worked beautifully. Sometimes the build broke at 5am and I would wake up to error logs.

The process was messy. There were em dash hunts across every page because I kept finding them in body text. Cached pages that would not update no matter what we tried. Fights about delegation when Clarence tried to make decisions I had not approved. At one point, Clarence ran on the wrong model (Sonnet instead of Opus) for an entire day because a settings file could not be edited from the interface we were using. The output quality dropped noticeably and we did not figure out why until that evening.

The part that matters most: Clarence has a knowledge database with over 1,500 memories built from our conversations, my coursework, and every correction I have ever given it. It distills what I care about, how I think, and what I will not tolerate. The portfolio content was not generated from a generic prompt. It was extracted from a system that knows my voice, my work, and my standards because I spent months teaching it. That database is the real artifact. The site is just what it looks like when you render it.

What Each Side Brought

James

Vision and creative direction
Taste. Knowing when something looked wrong even if it was technically correct
Domain expertise in UX, music, and the subject matter of every case study
Quality control from a real device in real lighting conditions
The word "no" when Clarence got ahead of itself

Clarence

Execution speed. Parallel agents building multiple pages simultaneously
Memory systems that tracked every decision and preference across sessions
Code generation, build management, deployment pipeline
The ability to hold the full context of a growing codebase in working memory
Patience for the fifteenth round of "make this two pixels smaller"

What is Honest About This

The portfolio content comes directly from my grad school coursework at Kent State. Every case study is grounded in real academic work, not made up for a portfolio site. The images are still missing. It is text-only right now. That is honest, not a failure. The words came first because the words are the work.

The changelog tracks every commit and who made it. This is not a "look how amazing AI is" story. It is a "we built something real with a lot of friction and it works" story. The friction is the interesting part. It is where the design questions live.