Every problem has a solution.

Clarence is a self-hosted multi-agent memory system on a single SQLite database: structured memories, a knowledge graph of entities and facts, semantic vector search, scheduled jobs, and public telemetry. Only one agent writes to memory. Every other agent submits notes that it accepts or rejects.

I tested search with 40 questions against the live database. Adding a reranking step, which re-sorts the results, put the right answer first on 34 of 40, up from 19. Reranking shipped as an optional add-on, off by default.

The main open problem is visibility of system status, Nielsen's first heuristic: Clarence can fail without telling me.

SensorSynthFM is an iPad FM synthesizer that treats motion, touch, and room energy as modulation sources. It is my Kent State MS UX capstone, built in Swift with AudioKit and SwiftUI.

The September 20 build passed 57 Xcode tests and my review on a physical iPad. Held chords, retriggering, and removing single notes all worked. Outside users have not tested it yet, and recruitment for moderated sessions began in November. I have not confirmed the ten-voice target.

Clarence writes code fast. It cannot decide whether a mapping feels musical. The musical and design decisions are mine.

Focus Reader is a free reading app I am building for iPad and iPhone. On the iPad I draw across a sentence with an Apple Pencil, and on the iPhone I touch, hold and drag. The highlight becomes a card with a citation and page number.

It reads PDF, Word, Markdown and text files. Markup saves into the PDF, and notes export as Markdown, a plain-text format for notes.

It is the first working piece of Focus, a system I designed in 2024 to keep what you read and connect it. The first build ran on my iPad on October 31, 2026. Claude wrote most of the code to my design, and I tested each build.

It runs on my devices through TestFlight, Apple's system for testing apps before release, and is not in the App Store yet. Still to come: the reading dashboard, EPUB books, web articles, the Focus modes from the 2024 prototype, and links between ideas across documents.

From 2024 to 2026 I wrote five studies of Ableton, 55 pages in all: a stakeholder analysis of the company and Live (2024), Nielsen heuristic evaluations of Note (2025) and Learning Music (2025–26, severity-scaled), a cognitive walkthrough of Learning Synths (2026), and a cognitive accessibility evaluation of ableton.com and Live 12 (2026).

Two problems repeat across Note, Learning Music, and Learning Synths. Each assumes knowledge a beginner does not have, and each expects users to go in order. Note has no in-app help and Learning Music has no glossary.

Live 12 added screen reader support, keyboard navigation, and high-contrast themes. I found nothing in it aimed at cognitive barriers. The accessibility report makes seven recommendations, starting with text labels on icons, which Push hardware already has.

I co-founded the Pittsburgh Ableton User Group in 2015 and have led it since, serving as direct liaison to Ableton's community team. I have watched hundreds of beginners use these tools for the first time.

The result is a VPAT 2.5 accessibility conformance report. I tested with axe DevTools, WAVE, HeadingsMap, and my own Python static-analysis scripts. Three criteria are documented Does Not Support: 1.4.1 Use of Color, 1.4.3 Contrast (Minimum), and 4.1.2 Name, Role, Value. axe found 74 issues on the NPR home page and 7 on a story page.

My Ableton evaluation follows COGA, the W3C guidance on cognitive and learning disabilities. It covers ADHD, dyslexia, working memory, and executive function. Next: open-sourcing an audit-to-VPAT reporting tool built from the NPR scripts.

This page has visible focus on every control, honors reduced motion, has a button that pauses motion, and needs no interaction to reach portfolio information.