Thesis

Constrained Evolution

Environmental Pressure, Sovereign Computing, and the Convergence of Biological and Computational Systems

Author: the ecoPrimals project Status: Working draft — all 16 chapters transplanted, undergoing refinement License: AGPL-3.0

Historical snapshot. Metrics in the thesis body reflect the gen3 working draft (circa February 2026: ~757K LOC, ~104K tests, 11 primals, 5 springs). Current ecosystem metrics: 3,598,358 LOC, 135,000+ tests, 15 primals, 9 springs (Evidence Snapshot). The thesis numbers will be reconciled as chapters undergo refinement.

Thesis Statement

Strategic environmental constraints — in both biological and computational systems — do not merely accelerate convergence to known solutions. They reshape fitness landscapes, driving specialization toward constraint-specific optima through independent evolutionary trajectories. This principle, observed in thermophilic adaptation, controlled laboratory evolution, natural population genomics, and AI-assisted software development within a strong type system, constitutes a general theory of constrained evolution applicable across domains.

This thesis presents the theory, the computational system built under its principles, and the empirical scientific validation proving the system computes real physics, biology, chemistry, and mathematics correctly.

Structure

Part I — Foundations

0. Front Matter — Abstract, acknowledgments, and dissertation metadata
1. Introduction — Taq polymerase motivation, thesis statement, five contributions
2. Literature Review — Extremophile biology, LTEE, type theory, AI-assisted development

Part II — Theory

3. Theoretical Framework — Formal constrained evolution: fitness landscapes, biology→computation mapping, predictions
4. Accept and Generate — Nature’s strategy for hard problems — generators, verifiers, enzymes

Part III — The System

5. System Architecture — ecoPrimals sovereign platform: primals, compositions, NUCLEUS
6. BarraCuda — Vendor-agnostic Pure Rust GPU compute (WGSL/Vulkan, f64)

Part IV — Experimental Validation

7. Experimental Methodology — The spring framework: phased validation across domains
8. Results: hotSpring — Computational plasma physics — Sarkas MD, nuclear EOS, lattice QCD
9. Results: airSpring — Precision agriculture — FAO-56 ET, sensor calibration
10. Results: wetSpring — Life science & analytical chemistry — 16S, QS, phylogenetics, PFAS
11. Results: groundSpring — Measurement noise & uncertainty — the tolerance foundation
12. Results: neuralSpring — ML primitives, Isomorphism Theorem, coralForge

Part V — Analysis

13. Quantitative Evidence — NTT→FFT evolution, convergent IPC, fastidious specialization
14. Biological Validation — LTEE frozen-fossil sequencing proposal

Part VI — Synthesis

15. Discussion — Strengths, limitations, alternative explanations, trade-offs
16. Conclusion — Five contributions, future work, closing synthesis

Back Matter

• A. References — Full bibliography
• B. Appendix: Hardware Inventory — See Contact
• C. Appendix: AI Methodology — See Sharing the Pen
• D. Appendix: Spring Validation — See Spring Catalog

How to Read This

If you are a committee member: Start with Introduction for the thesis statement, then Theoretical Framework for the core argument, then Methodology and any results chapter in your domain.

If you are evaluating the science: Start with Methodology and the results chapter for your domain. Each spring is a self-contained validation study with public repositories you can clone and run.

If you are evaluating the system: Start with System Architecture and BarraCuda, then Quantitative Evidence for the NTT→FFT constrained evolution case study.

If you are interested in the biology: Start with Theoretical Framework sections on Taq, Lenski, and Anderson, then Biological Validation for the LTEE sequencing proposal.

Lineage

This thesis evolved through three generations:

1. Inoculum — constrained_optimization_ai.md — first formulation during the ecoPrimals build. Rough metrics, illustrative analogies, no empirical validation.

2. Working Paper — Constrained Evolution — Formal — reframed from “optimization” to “evolution” after engaging with the biological literature. Added Lenski LTEE, Taq, Anderson genomics, and the firefly/symbiotic composition arguments.

3. Dissertation — this section — full academic work with literature review, formal mathematical framework, five chapters of empirical results, proposed biological validation, and honest discussion of limitations.

The philosophical counterpart lives in atlasHugged. Where the thesis asks “does constrained evolution work?”, atlasHugged asks “why does it matter?” — in particular, The Human Search is the “napkin version” of Chapter 03.

A Note on AI-Assisted Writing

This thesis, like the system it describes, was produced with AI assistance (Cursor IDE, Claude). The Methodology chapter documents this explicitly. The AI is the mutation operator; the builder provides constraint and selection. The thesis is itself a product of the constrained evolution methodology it formalizes.

The public spring repositories contain runnable experiments that verify every quantitative claim. The science stands independent of how it was written. See Sharing the Pen for the full methodology.