Research notes
Essays on markets, risk, and learning systems. Each note links to the full text.
How Agents Run the Research PipelineHow research rules become contracts that agents can execute, why every rejection is a first-class output, and which decisions must stay with people.Two Languages of Risk ControlSizing decides how large a bet should be; safety decides when to stop. They are different control systems, and merging them is the most dangerous mistake in risk design.The Geometry of AlphaWhat compounds is not a recipe but the factory that keeps producing, validating, and retiring signals, and the geometry that explains why many weak, independent signals beat one strong one.Chern's Question, Answered by Two WorldsA question Chern left behind about piecewise manifolds, later answered independently by rock mechanics and by neural networks.The Four Realms of Neural NetworksFour lenses on deep learning: PDE solvers, manifold geometry, gauge fields, and quantum-like attention. Each reveals structure the one below cannot express.Why Backtests Fail in Live TradingSeparating signal failure from fill assumptions and accounting errors, so the gap between backtest and live is explained rather than blamed on costs.