paper: main.tex
note: >
  Scripts reproduce the computations reported in the paper. They are evidence, not proofs.
  Requirements: Python 3, numpy, scipy, pycosat (SAT rows). Run from inside anc/.
rows:
  - claim: "Echo Lemma, song 01, every width <= 34"
    script: certtree.py
  - claim: "Echo Lemma, songs 001 and 0001, every width <= 20"
    script: certify_all.py   # uses masked_sat.py (pycosat)
  - claim: "Echo Lemma, song 011111, every width <= 80; stall statistics"
    script: stall3.py
  - claim: "Two-sided, alternating centre, every seed of radius <= 10; max (T - R) table"
    script: actual.py
  - claim: "Two-sided, right radius <= 14 and left radius <= 125"
    script: rightattack.py
  - claim: "Merging certificate, song 011111, M = 76"
    scripts: [merge.py, independent_merge_check.py]
  - claim: "Forced identities A2..A7 checked on exact forced spacetimes; failure of the propagation law at depth 9"
    script: idcheck.py
  - claim: "Constant half-columns up to depth 41; no-domino band"
    script: zerocols.py
  - claim: "Forced creation E = d(t+2) under the echo"
    script: ebit2.py
  - claim: "Rule 150 self-similarity identity; Rule 30 comparison (50.2%)"
    scripts: [r150.py, selfsim30.py]
  - claim: "No additive conserved quantity on patterns of length <= 5"
    script: energy.py
  - claim: "Closed global parities are memoryless"
    scripts: [twonight.py, closed2.py]
  - claim: "Distinguishable sideways histories (lower bounds)"
    scripts: [mn2.py, window.py]
  - claim: "16-gate classification; pinned-zone machine and layers"
    scripts: [gates.py, pin.py, machine.py, layer3.py]