Methodology
How it was built — constrained evolution, K-NOME programming, and the operational playbook.
The methodology is biological: evolve under constraint, validate against published science, compose from small parts, track everything. Two ideas drive everything else — constrained evolution (remove dependencies, force genuine capability) and K-NOME (AI as collaborator under human constraint, every generation tested against published results).
- Constrained Evolution — Formal — why removing CUDA produced vendor-independent GPU compute, and other constraint-driven innovations
- K-NOME Programming — the operational model: human domain expertise + AI implementation, every generation validated
- How to Start a Spring — the phased playbook: Python → Rust → GPU → composition
- Knowledge Commons Targets — 9 domains where public data + cheap hardware unlocks sovereign alternatives
- scyBorg Licensing — AGPL + ORC + CC-BY-SA: three independent nonprofits, no single entity can revoke
- P vs NP and the Enzyme Thesis — why generation/verification asymmetry matters for computation and biology
- Constrained Optimization in AI-Assisted Development — the initial formulation: constraints + direction + iterative AI = rapid convergence
- Constrained Evolution — Thesis — the full academic argument: 16 chapters, 9 springs, LTEE sequencing proposal
- Sharing the Pen — operational model for AI-assisted constrained evolution