The Quiet Threat to Every Frontier Model Lab: Knowledge Distillation
How enterprises are quietly exiting frontier model pricing, and how Chinese labs are using the same technique to extract capabilities at industrial scale. The technique is ancient. The stakes are new.
→Every True Atom Must Earn Its Truth — A Deep Dive into ASP
A ground-up explanation of Answer Set Programming: why absence means false, what stable models really are, how grounding works, and what separates ASP from MIP, CP, and SAT. Built carefully from first principles.
→GPU for Constraint-Based Search — Literature Map
A literature map charting papers on using GPU parallelism to accelerate CP solvers, SAT/SMT solving, branch and bound, and local search. Annotated as I read them.
→The World of Scheduling Problems
From factory floors to ride-hailing — how mathematics orders the chaos of competing tasks across limited resources in time. An interactive guide to 8 problem families, complexity theory, and real-world applications.
→Scheduling with CP-SAT — A Hands-On Tutorial
From a single machine to RCPSP — how to model and solve real scheduling problems in Python using Google's CP-SAT solver. Animated Gantt charts, working code, and intuition at every step.
→When Scheduling Meets Reality — Vertical Integration
The bottleneck is no longer solver performance. It is translation — from real-world complexity to deployable decisions. Five challenges, five opportunities, and why the next generation of decision platforms will be judged by how well they integrate.
→Scheduling in the Wild — Industry Applications
How six industries tame their own version of the scheduling problem. Manufacturing, healthcare, aviation, logistics, retail, and project management — each with its own constraints, failure modes, and solver approach. Includes two case studies: UPS ORION and NHS nurse rostering.
→ARC Consistency — From CPU to GPU
My attempt to redesign arc consistency for the GPU — what I had to rethink, where the algorithm resisted, and what I learned along the way. Not a performance study, just an exploration of what changes when you move a sequential idea onto parallel hardware.
→How much can you get from your CPU — and when do you need a GPU?
Most Python code uses a fraction of available CPU power. An interactive explainer on parallelism, OpenBLAS tiling, and where the CPU ceiling actually is.
→Stock Market Series — A Guide I Made for My Son
He came to me with questions about stocks and how people make money from investing. An interactive, quiz-based series — Beginner and Intermediate — starting from zero and building up progressively.
→The GenAI Wall — why domain expertise is the real bottleneck
Coming soon
What cuOpt actually does — and what it doesn't
Coming soon
Stochastic optimisation — one model or many solves?
Coming soon
We Build More Than We Use
Creation has become easier. Adoption has not. On the gap between output and absorption — and why knowing the difference is becoming a core skill.
→The GenAI Landscape — structured
Foundations, architectures, tooling, and production patterns across modern Generative AI — navigable as a structured mindmap.
→Scheduling with CP-SAT — A Hands-On Tutorial
From a single machine to RCPSP — how to model and solve real scheduling problems in Python using Google's CP-SAT solver. Working code, animated Gantt charts, and intuition at every step.
→Stock Market Series
Two-level interactive guide — from what a stock is all the way to financial statements, technical analysis, and building a real portfolio.
→