Writing
Essays about claims, tools, and the math hiding under both.
The blog moves between AI systems, local-first software, programming correctness, and the small tools that make technical work less hand-wavy.
The Math Remembers: How to Catch a Model Lying About Being Original
Model claims are cheap. Weight arithmetic is harder to fake. A field guide to merge forensics and its limits.
LLMmodel forensicsmachine learningJune 2026Bounded Wrongness: A Field Guide to Probabilistic Data Structures
Bloom filters and their relatives trade exact answers for smaller memory. The trick is knowing which kind of wrongness you can afford.
algorithmsdata structuresdatabasesJune 2026Your Music Has No Genre: Reading It Off the Waveform
A practical turn away from metadata lookup and toward local audio analysis for people with messy music libraries.
audio MLlocal-firstmusic techJune 2026Your Coding Agent Has Amnesia: It's the Harness, Not the Model
A good agent is not just a model. The harness, memory, and handoff structure decide whether it can keep its head across a long task.
coding agentscontext engineeringdeveloper toolsMay 2026Memory as Polynomial Projection
HiPPO, S4, Mamba-style models, and the difference between long context as storage and long context as prediction.
state space modelsdeep learningtime seriesJanuary 2026JavaScript Date Is Broken. Here's What Replaces It
Temporal makes date and timezone bugs harder to write. The post pairs the argument with a scheduler implementation.
Temporal APITypeScriptweb dev