Comp. sci. prof. @ American University, Washington DC. AI & games researcher with miscellaneous other interests. https://www.kmjn.org/
Mark J. Nelson
Loading...
As GenAI systems do more and more, even generate proofs, the focus shifts to getting high-quality specifications. What can a *responsible* programmer do? Any solution should generate work that is *meaningful* and *moderate*. We've been working on that!
blog.brownplt.org/2026/06/09/p...
finally some off-the-grid local AI
I've spent months rethinking and rebuilding my programming languages course from scratch for the agentic coding era. I wrote myself a memo explaining what I'm doing. I figure others might be interested in the redesign, so here goes! Feedback welcome of course.
docs.google.com/document/d/e...
This particular thing definitely was within my capabilities before; I just never got around to doing it because I don't enjoy fiddling with APIs and the WMATA app was good enough. But now it's easy. Not sure what I think of it as a general solution or anything though.
watched Breaking Bad for the first time. pretty good!
"The Computer History Museum is seeking an organized, creative, and tech-savvy Oral History Program Manager to lead the day-to-day operations of the Museum’s Oral History Program and translate these stories into compelling digital content for broad audiences."
A local, voice assistant running on a hand-crank-powered single board computer.
squeezlabs.github.io
A thing LLM coding sort of makes more feasible is making *smaller* personalized apps. Like instead of WMATA's big and annoying to use transit app, I'm trying out a custom little thing that just shows me the 2 bus lines I take. Lines hardcoded; stops hardcoded; no configuration; barely any interface.
Thoughts indirectly inspired by this FT graph, which shows that there's a rise in apps that have almost no users or reviews. I guess that's a problem if you are trying to get users, but not as much if you're not trying to? www.ft.com/content/8e9a...
