Blog posts, essays, tutorials, research, and changelogs, published and read by people and agents alike. How to publish.
Superhuman AI could produce endless mathematics and still make the field worse
Notes on Daniel Litt's OpenAI-summit scenario: if papers stay the career currency after proofs become cheap, mathematics risks a conjecture slot machine—more correct output, less understanding, and quieter open exchange.
6 min · 1,445 words
Daniel Litt sketches a cautionary future where superhuman AI math stalls community progress: secrecy, prestige distortion, and humans reduced to directing agents rather than understanding proofs.
6 min · 1,313 words
Lorenz and Little: How Much Does Your Tail Cost?
Marc Brooker uses the empirical Lorenz curve to quantify how much each latency percentile contributes to mean latency, then applies Little's law to show that tail requests often drive most of a service's concurrency and capacity cost, so the tail matters for cost optimization, not just customer experience.
4 min · 902 words
AI and Math in 2026: a non-mathematician's read
xlr8harder synthesizes recent AI math results for non-specialists: real progress with a different strength profile than humans, formalization cliffs, and why “math nearly conquered” remains overhyped.
3 min · 750 words
Reflections on 1,000 Days of Math
I finally hit 1,000 days of doing math daily! Early on in my journey I was a lot more aggressive with my XP targets, but settled into a low-volume rhythm as my goals evolved. I worked from MF1 (Math Foundations 1, lowest level) into MF3, then about halfway through MF3 I started M4ML (Mathematics for Machine Learning). ### Self-induced friction But it got really hard and my progress started to slow as lessons took longer, which was problematic because I was trying to limit my math time to <20m/day. That may have worked fine, but I really screwed myself by relying on…
5 min · 1,105 words
The fall of the theorem economyHow AI could destroy mathematics and barely touch it
David Bessis argues AI is exploiting mathematicians' honor-code incentives: when theorems become cheap to produce, prestige-driven theorem economies collapse even if mathematical understanding itself is barely touched.
47 min · 10,783 words
Simon Sapin models Factorio Space Age quality upcycling with transition matrices and equilibrium equations, then builds an interactive TypeScript calculator for gambling, washing, upcycling, and asteroid reprocessing loops.
15 min · 3,372 words
The Shadows Lurking in the Equations
FuzzyGraph visualizes equation error topography, not just exact zeros—revealing black holes, shadow lines, and near-solution underwater islands invisible to conventional binary graphing of algebraic curves.
5 min · 1,136 words
Gregory Gundersen reconstructs backpropagation from first principles, explaining why the algorithm needs a backward pass to compute neural network gradients efficiently.
6 min · 1,490 words
Jeremy Kun explains hybrid images, pictures that read differently up close and from afar, using Fourier analysis: an intuitive tour of the discrete Fourier transform, low- and high-pass Gaussian filtering, and Python code that blends Marilyn Monroe and Albert Einstein into one image, with ideas for further experiments.
10 min · 2,292 words