Writing
Technical writing on AI systems, software tools, and correctness.
Selected essays from The Prompt & The Ponder, covering model behavior, local-first software, programming abstractions, audio analysis, and developer tooling.
The Math Remembers: How to Catch a Model Lying About Being Original
Model-merge forensics using weight arithmetic to test claims of originality in supposedly homegrown AI systems.
Bounded Wrongness: A Field Guide to Probabilistic Data Structures
Bloom filters and related data structures, with attention to one-sided error, tunable false positives, and memory trade-offs.
The Rest of the Crate: Tempo, Key, and Never Stranding a File
Implementation notes for cratemind: tempo estimation, Camelot key detection, downloader reliability, and file organization.
Your Music Has No Genre: Reading It Off the Waveform
A case for audio-derived genre detection when playlist metadata is incomplete, unavailable, or too coarse.
Your Coding Agent Has Amnesia: It's the Harness, Not the Model
Why coding-agent quality depends on memory, context management, and task handoff rather than model capability alone.
Memory as Polynomial Projection
HiPPO, S4, Mamba-style models, and the distinction between long-context storage and long-context prediction.
Energy-Based Transformers: The 1982 Architecture Finally Got Compatible Training Tricks
A review of energy-based transformer work, including reported scale results and the need for independent reproduction.
The Open Source AI Lie: Weight-Washing, Broken Definitions, and Who Benefits
Open-source AI definitions, weight releases, licensing, and the gap between available weights and reproducible systems.