PACSeries v0.2: Dawn Field Theory Six preprints developing a single idea: that information erasure creates structure, and structure obeys a conservation law. The conservation law is PAC (Potential-Actualization Conservation). Its unique stable solution is the golden ratio. The rest follows. What's in the papers: Paper 1 starts from Landauer's principle and the data processing inequality. Both are textbook results. Combined, they require that erasing information into a multi-mode environment creates correlational structure between the modes. The collapse efficiency ratio converges toward ln(phi) at 0.15%. Paper 2 shows that five independent computational domains (Fibonacci arithmetic, cellular automata, prime sieves, Landauer erasure, and Mobius field dynamics) all produce a balance constant near 1.058. This decomposes as gamma + ln(phi), where gamma is the Euler-Mascheroni constant. The probability of this clustering by chance is p < 0.0003. Paper 3 presents closed-form expressions for the Feigenbaum constants using only pi, Fibonacci numbers, and small integers. The accumulation point r_infinity matches to 13 significant figures. No causal claim is made. Paper 4 writes Standard Model parameters as Fibonacci ratios. The fine-structure constant matches to 5.7 ppm. The weak mixing angle is 3/13. A falsifiable prediction is included: a Z' boson at 395 +/- 20 GeV. Paper 5 derives Maxwell's equations and three spatial dimensions from information-theoretic constraints. Five independent arguments each require exactly D = 3. Paper 6 shows that PAC conservation appears in trained neural networks without being programmed in. SEC phase classification predicts token accuracy with zero free parameters. Hallucination corresponds to uncompensated entropy. What's in the package: Each paper includes the full text (Markdown, LaTeX, PDF), numbered experiment scripts, raw JSON results, and publication figures. 76 experiments total. Everything runs with standard Python (NumPy, SciPy, Matplotlib). A summary paper is included for quick orientation. Honest caveats: These are preprints, not finished papers. Papers 1-3 rest on established mathematics. Papers 4-5 make precise numerical matches but the theoretical grounding for why PAC recursion constrains gauge structure is interpretive. Paper 6 is validated on small models only. Every paper states what would falsify it. Failed predictions are documented where they occurred. What's next: The current papers are being tightened for formal submission. Specific areas of ongoing work: Strengthening the gamma interpretation in Paper 2 (currently gamma and 1/sqrt(3) perform comparably as decomposition candidates) Extending Paper 6's neural network validation to larger models Independent replication of Paper 3's Feigenbaum formulas New experimental work is underway (milestone 4, early stage) exploring three directions: PAC Relativity: whether the Lorentz factor is a PAC energy partition between internal cascade and propagation. Preliminary results show exact mathematical equivalence, but analytical formalization is needed. Turbulence from cascade mechanics: whether Kolmogorov -5/3 scaling emerges from PAC cascade at physical mode counts. Current best result is -1.612 (3.3% from -5/3) at 8 modes. Energy as collapsed potential: connecting nuclear binding energies and configuration space to the Landauer cascade framework. These are exploratory and may or may not become papers. Results will be published as they mature. Previous version: v0.1 (October 2025) contained 5 papers and is available as Zenodo record 17295103. v0.2 is a complete restructure with new experiments, tighter statistics, and three new papers. Code: https://github.com/dawnfield-institute/dawn-field-theoryLicense: AGPL-3.0