by Daniel Litt
Daniel Litt, an algebraic geometer at the University of Toronto, is the most clear-eyed mathematician on X about AI: he tracks what models can and can't do in research mathematics and argues it with specifics rather than vibes. In between he posts short, lucid sketches of real mathematics, like a recent thread on where Deligne's proof of the last Weil conjecture fits.
March 2, 2026
Erik Hoel on why we can't trust AI self-reports, Peter Singer's moral evolution on Philosophize This!, and the Kurzgesagt video that made infinity feel small.
February 9, 2026
Nine mathematicians prove the geometric Langlands conjecture, 3Blue1Brown visualizes attention in transformers, and CGP Grey declares war on the nickel.
January 5, 2026
Matt Yglesias grades his 2024 predictions, Tyler Cowen spots AI in the productivity data, and Numberphile finds a new recipe for an old number.
