Brutus and I Went to Indianapolis

I presented a poster asking how much of a Transformer’s apparent reasoning is actually learned. Indianapolis, August 2026, at a conference where Chomsky’s branch of NLP turned out not to be dead.

A third time in Philly, for the mathematicians

July 2026, the International Congress of Mathematicians in Philadelphia, the first time in nearly half a century it has assembled in the United States. Four Fields Medals, an itinerary built by an agent, and a Chern lecture on formalising quantum field theory.

I skipped CVPR for an interferometer in a box

A week in the Research Triangle instead of Denver: the Quantum Machine Learning Workshop at NC State, the Duke photonics lab over in Durham, and a teaching kit that packs a whole interferometer into a box.

A porch in San Diego, and an entire arm of talks about creativity

December 2025, NeurIPS in San Diego. A year earlier I skipped it entirely and spent the week arguing with a text to video model instead. This time I went, and the thread I could not shake was creativity, and the writers who want nothing to do with any of it.

AI for the Accelerated Science, downtown

The Chen Institute’s AI for the Accelerated Science symposium in San Francisco, October 2025.

NeSy25 in Santa Cruz, and a Mario Kart run over HWY17

September 2025, NeSy25 in Santa Cruz, where the whole emphasis was on making models reason and then show their work. I also took HWY17 over to the Imagination in Action event at Moffet Place, and missed Neal Stephenson for my trouble.

Two Tracks in the House That Halmos Built

A conference split cleanly into Quantum for AI and AI for Quantum, plus an NVIDIA roadshow aimed squarely at midwestern researchers. Indiana University Bloomington, 2025.

Ed Chi on machine dreaming, and a vibe shift at Berkeley

August 2025, the Agentic AI Summit at UC Berkeley, the middle stop of nine days and three towns. A sidebar with Ed H. Chi of Google DeepMind on why models dream, the inductive-deductive divide, and why both sectors got to the same idea at the same time.

The Trillion Parameter Consortium, and a Vibe Shift

National-lab researchers working out how AI might not just assist large-scale physical science but autonomously conduct it. Santa Clara, July 2025, the first of three towns in nine days.

Motor City to Music City, and the hat they ran out of

June 2025, CVPR in Nashville. I came in with a list of workshops and ended up filing from the floor: fal out of GPU Rich hats, Lambda’s consumer boxes, a LIDAR bolt-on kit at the Latitude AI booth, and one evaluation that did not go the way I expected.