# The three golden paths — cold-run records

Each path was run cold on 2026-11-16 from a fresh copy of the collection
tree, in a clean directory, with no author state. What follows is what
actually happened, including the friction.

## A · OmnaScript — PASS
- obtain: kernel-collection → languages/omnascript (13 files in its node tree)
- run tests: `node omnascript.test.cjs` → **pass 21, fail 0**
- compile a specimen: `node omnascript.cjs examples/island.omna --out island.compiled.cjs`
  → compiles; the compiler files source and compiled sha256 itself
- known limitations: the collection copy carries 21 tests; the consolidation
  tree's larger suite (36) is the fuller record. The language declares;
  a JavaScript executor carries — see the entity's commonMisinterpretations.

## B · SlangLang — PASS
- obtain: kernel-collection → languages/slanglang
- run tests: `python3 test_slanglang.py` → 67 tests OK;
  full runner `PYTHON=python3 node run-tests.cjs` → **Ran 78 tests — OK**
- friction found: the runner refuses to guess a Python executable
  ("Set PYTHON to a trusted Python 3.10+ executable") — set PYTHON first.
  That refusal is deliberate design, and now it is documented design.
- the published claim (78 tests green) reproduces to the exact number.

## C · Harakiri NG — BUILD PASS · QEMU RE-RUN OWED
- obtain: kernel-collection → kernels/harakiri-ng-zig
- build: `zig build` (house Zig 0.14) → succeeds cold;
  **harakiri-ng.bin = 531,200 bytes, identical in size to the proven
  v1.1 baseline** recorded in BUILD-CARD-v1.1.md
- verify procedure (extracted from the tree, canonized): build, then
  `./v10_regression.sh` — boots the .bin in QEMU (qemu-system-aarch64,
  virt/cortex-a72) and runs every PoeTchi REPL command once; 16 checks.
- tonight: the regression was NOT re-run — qemu-system-aarch64 is not
  installed on this machine. Its last full run is filed inside the tree:
  test/poetchi-v11-regression-2026-11-02.log, 16/16 PASS, 2026-11-02.
- there is no `zig build test` step; the regression script is the test door.
