James Dishman

I built an autonomous AI system that manages a knowledge base of 3,397 active memories, 2,474 entities, and 10,349 active facts with full vector search across 2,075 indexed vault notes, routes tasks across multiple models by cost and capability, syncs reference materials from Google Drive, and posts reports to 9 Discord channels. Then the platform I built it on changed the rules, and I had to migrate the entire system under constraint in 48 hours. The real lesson was not about automation. It was about trust calibration, platform dependency, and what happens when you design a collaborator instead of a tool.

Multiple UX evaluation methods applied to three Ableton products, grounded in deep music production experience and a decade leading the Pittsburgh Ableton User Group. The suite revealed a consistent pattern: Ableton designs for insiders, and that works until someone shows up for the first time.

I ran a multi-method accessibility audit across music tech, news, social media, and academic sites. The work covered WCAG compliance, hands-on VoiceOver screen reader testing on GroundNews, an Instagram accessibility audit of a university account, contrast analysis, PDF remediation, and cognitive accessibility research. The biggest finding was not any single violation. It was that accessibility best practices for people and machine readability for AI are the same goal, and most teams are failing at both.