Essays
Long-form pieces on markets, systems, and the marginal cost of reasoning. Many apply the principles behind custom decision engines— repeatable logic, clear criteria, and defensible numbers. Drafts may evolve.
Definitive guides to specific problems, not just updates. If you want to go deeper on error handling, TCO, or evaluation frameworks, start here.
Why the System Feels Rigged: The Balance-Sheet Truth
The widespread intuition that the economic game is rigged isn't an emotional illusion or bad luck. It is the direct mathematical output of modern double-entry plumbing. Here is the balance-sheet proof.
PhaseSAE: Time-Division Multiplexed Dictionary Learning for Memory-Efficient Mechanistic Interpretability
PhaseSAE introduces Time-Division Multiplexing (TDM) to sparse dictionary learning. By streaming P orthogonal parameter slices against progressive residuals, PhaseSAE strictly reduces peak memory bandwidth to 1/P while boosting downstream Cross-Entropy recovery by +65.9% on TinyStories-1M.