Articles
Longer notes on AI, trading infrastructure, correlation, causation and control.
Flagship
Correlation, constraints and control
AI, trading systems and blood glucose control studies share a common structure: statistical inference is useful, but control and stability require explicit constraints.
Draft placeholder
Correlation and causation in large language models
Why it is both true and incomplete to describe LLMs as correlation engines.
Draft placeholder
Training AI on FIX message flows
Representation, confidentiality, regime change and validation when applying machine learning to structured trading logs.
Draft placeholder
Compression as a form of representation
FAST encoding, model training and the relationship between compact representation and useful structure.