When the Bug Is in the Instruction
A bug in agent-written code is often not a bug in the code. Because the build prompt lives on the issue, you can debug the instruction the same way you debug the program.
agenticai-toolsworkflow
AI-E 
I work at the point where emerging technical capability becomes operational reality. Research, enterprise systems, open source, teaching, and now AI — the recurring pattern is the same: identify, build, help adapt.
Teaching
Teaching courses in data science, data and software engineering, with substantial GenAI integration, at Boston University.
Working on
Exploring where institutional capacity meets agentic systems.
Operating the AI-Native Stack
Essays & Writing
A bug in agent-written code is often not a bug in the code. Because the build prompt lives on the issue, you can debug the instruction the same way you debug the program.
By every metric a distributed-systems engineer cares about, a GitHub issue tracker is a terrible message bus. The properties that make it terrible are exactly the ones agentic workflows need.
My Calendly trial lapsed and the booking link on my site started 404ing on visitors. I was on Cal.com 30 minutes later. The interesting part isn't that the AI wrote code, it's what it did to the switching cost I'd stopped questioning.
Three of many
A four-level taxonomy for describing how AI was involved in a piece of work — built for journalism, academia, and professional practice, and positioned as an open standard.
Methodology and program design at the intersection of journalism, computation, and institutional practice — including TRACE, open tools, and curricular experimentation.
Systematic protocols for integrating generative AI across data science coursework — balancing productivity with educational integrity and professional responsibility.
Semantic Trails
Trails are author-curated paths through the body of work — linking essays, older projects, and material that wouldn't surface in a feed. Start anywhere and follow the connections.
A trail on ambient presence, consulting infrastructure, and the things remote work tooling keeps getting wrong — and what that reveals about what presence actually requires.
2 stops