Course
The Periodic Stack
10 lessons. ~16 minutes. From Landauer to silicon.
Each lesson pairs with an interactive view in the instrument.
Landauer's Principle
Why erasing a bit of information has a physical cost. The thermodynamic floor of computation. kT ln 2 per erased bit — non-negotiable.
Read the lesson →Bennett's Reversibility
How any computation can be made reversible with enough ancilla bits. The L₀ vs L₂ split. 26 Bennett decompositions in the stack — each one names the free kernel inside an ostensibly paid operation.
Read the lesson →The 6 Axes
Family, abstraction, thermo cost, concurrency, state topology, information topology. Six orthogonal axes. Every operation has a position. The position tells you the physics.
Read the lesson →The 33 Families
From foundational algebra to noncommutative geometry. A tour of every algebraic family that has demonstrated computational character. 17 from v1. 16 more from the 2026 mathematics survey.
Read the lesson →The 297 Edges
Composition: X is built from Y. Bennett decomposition: X = free phase + paid phase. Thermodynamic inheritance: child's cost is forced by parent. 343 edges turn the table into a directed graph — and the graph is a function.
Read the lesson →The Math Projection
The compute stack has 6 axes. Project away the 3 physics-dependent axes (abstraction, thermo, concurrency) and you recover the Math Periodic Stack — a purely mathematical classification of operations. 258 primitives, 33 families, 528-cell coordinate space.
Read the lesson →The Mendeleev Predictions
528 cells in the math tensor. Only 28% are occupied. Empty cells with occupied neighbors are predictions — operations the structure says should exist but nobody has named. The same logic that predicted gallium, scandium, and germanium.
Read the lesson →The Silicon Corridor
Apple, Google, Groq, AMD, Intel, Tenstorrent — all converge on the same 29 of 258 primitives. That's 11% of the stack. 25 of 33 families have zero dedicated hardware. The corridor is a consequence of economics, not physics.
Read the lesson →The Leverage Leaderboard
Not all 229 hardware gaps are equal. #76 matrix decomposition has 11 downstream dependents — build it and you unlock Kalman filters, eigensolvers, interior-point methods, all of LQR. The leaderboard ranks every primitive by leverage.
Read the lesson →Building Silicon from the Stack
The recursive inversion function: f(goal) → gaps. Give it a computational target, it returns the ordered list of missing hardware from leaves up. From gap analysis to FPGA prototype to ASIC tape-out.
Read the lesson →