Thoughts on people, AI, and people + AI. ✌🏻 ...
A private, speech-responsive cellular visualizer for live presentations. It listens to the changing energy, brightness, texture, and cadence of a speaker's voice, then turns those qualities into environmental forces inside a Life-inspired field. demo / repo
I really don’t like the framing of “ synthetic users ” for the new wave of agents simulating user feedback, grounded in actual user data or not. I much prefer and have been advocating calling them “ simulations ,” which is much more accurate to what is actually happening, is a well established approach to prediction in the absence of direct data, and avoids the pitfalls of putting them so…
🚀 Launching a little side project today I hope people find useful. It's called ◼️ Shape ◼️ and it’s a site where people can learn about shaping AI model behaviors through interactive playgrounds. https://www.shape-models.com/ The site is split into sections to learn about the common ways people shape model behavior, play with examples using actual models, and save what you make (if you want). You…
I asked ChatGPT (5.5) to help me brainstorm some ideas of how theory of mind (my dissertation topic) could be more deeply integrated into frontier AI model development to improve model capabilities and usefulness. This is the output, which is...good. Really good. A lot of ideas to follow-up on. Integrating Theory of Mind into Frontier AI Model Development Executive summary Theory of mind (ToM)…
A researcher's job is to create situations of high legibility. Answering the question, "What is happening here with these humans?" is naturally unclear. A researcher needs to intervene in the situation (i.e. the study design) in a way that maintains the phenomena of interest while increasing its legibility. This legibility from the built situation then becomes a boundary object for communication…
Substack note
Subagented /sʌbˈeɪ.dʒəntɪd/ verb : When you get delegated work by an AI, yours or someone else’s Ah man Jeff’s agent asked me to urgently review one of its PR’s right before I was heading home. I totally got subagented! Substack note
Slop squint /slɑp skwɪnt/ noun : Lazily judging if a work deliverable needs to be actually good or if slop will do I had to write a business justification for additional budget and thought it was going to take all afternoon. But I did a quick slop squint and realized I could just one-shot it and be done by lunch. Substack note
Effort Etiquette /ˈɛf.ɚt ˈɛt̬.ɪ.kət/ noun : The practice of matching your production effort to your audience’s review effort when using AI. He sent us this complete slop that was probably a one-line prompt and expects us to review all 20 pages of it! The guy has no effort etiquette! Substack note
Frontier UX Research Circa May 2026