# The fresh check's own code

Written by a second model that saw the claims and our code but none of our reasoning, to break them.
Different methods throughout. Run from this folder; Python 3 with numpy, and a C compiler for one.

| file | what it checks | how to run |
|---|---|---|
| `vtaps.py` | Its own linearisation of Starlock's painterly pass (81 taps) from `reveal3d.js`, and H. | imported |
| `c1_exact.py` | The space-time formula, by a different exact method (every tap only shifts the residue (Ax + By + Ct) mod N; the error is integrated exactly over the N + 1 level intervals): 400 random lattices, three filters, boxes, decays, one frame, and the shimmer. | `python3 c1_exact.py` (`../out/independent-formula.txt`, seconds) |
| `floors_v.py` | The smallest response, the still floor, the box and decay floors, by direct minimisation. | `python3 floors_v.py` (`../out/independent-floors.txt`) |
| `c4_search.py` | Every (A, B, C)/N up to 64 levels, its own vectorised search; tests "best box-k = best still / k". | `python3 c4_search.py [Nmax]` |
| `c5_sim.py` | The practical dithers and the golden shift by direct simulation (blend over frames first, then the taps). | `python3 c5_sim.py [x0 y0]` (about a minute) |
| `c6_periodic.c` | Paint off, one pixel: every periodic 0/1 sequence up to period 26 through a decay, with its exact steady state. | `cc -O2 -o c6 c6_periodic.c -lm && ./c6 40 26` |
| `c6_flip.py` | Flipping every frame through decays; the box floor for one pixel, attained. | `python3 c6_flip.py` |
| `c7_window.py` | Our shimmer estimator on synthetic signals with a known answer: it found the window too short. | `python3 c7_window.py` |
| `c7_spot.py` | Our renderer, still against the half-step flip, scored by our estimator and exactly (both are periodic, so two renders do). | `python3 c7_spot.py ring 0` (all 8 cases: `../out/independent-pipe.txt`) |