Understanding how systems behave when everything is connected.
Correlace examines AI, trading systems and adaptive control through the shared lens of correlation, constraints and feedback.
Modern AI systems are effective because they learn statistical relationships at scale. Real-world systems are difficult because they also require determinism, auditability, bounded behaviour and control. Correlace sits at the intersection of those two observations.
Purpose
Educate by explaining how modern AI systems function in practice, including the distinction between statistical modelling and causal reasoning.
Influence by describing mechanisms, limitations and failure modes in domains where correctness, stability and auditability are required.
Incubate by developing ideas and experiments in financial systems, AI tooling and adaptive control problems such as blood glucose management.
The current focus
Correlation and model behaviour
How language models use statistical structure, why they appear to reason, and where that approximation breaks down.
Trading systems and infrastructure
FIX flows, message sequencing, latency, auditability, market structure and the practical use of models in production systems.
Adaptive systems
Glucose modelling, insulin action profiles and the broader problem of prediction under safety constraints.
Start here
Correlation, constraints and control
The flagship article linking AI, trading systems and blood glucose control through the common problem of prediction under constraints.
Notes on using AI
Short practical observations on language model behaviour, including implicit metaphors, default patterns and plausible but wrong answers.