Repo polish, after the testing campaign

Status: DONE 2026-08-04. The eight-program testing campaign is complete: eight programs across six axes, zero compiler/runtime defects, six tool-gate lanes, and the byte-for-byte proof. Outcome: TODO sweep clean, LICENSE/SECURITY/CONTRIBUTING present, README evidence section added, CI stays local-only by owner decision.

2026-08-04: the eight-program testing campaign is declared complete. Eight programs across six axes (data structures, bits, arithmetic, networking + shared-state concurrency, algorithmic depth, file I/O), zero compiler/runtime defects, six tool-gate lanes, and a differential fuzzer, and the byte-for-byte proof. The completed plans live at docs/internals/plan-tycho-sat-DONE.md and the six before it; the fresh clone that proposed a build tool (the last untested axis, systems I/O) is superseded by this polish turn -- that axis stays filed as the next program if testing ever resumes.

The rule from the program plans carries over: a finding becomes a phase only when a second, independent caller exists.

The work

The repo measures clean (1224 commits, 894 .ty files, ~9 grep hits that turn out to be false positives -- see phase 1). The polish gaps, in priority order: a LICENSE (the legal blocker), a GitHub Actions workflow running make ci, a README section documenting the seven tool programs as the evidence of exercise, and a CONTRIBUTING.md.

Phases

Phase 1 -- sweep the TODO/FIXME markers [DONE 2026-08-04]

The tree carries no TODO/FIXME/HACK markers at all. A whole-word scan of every .ty, .md, .c, .sh and .py file (including src/tychoc.c and runtime/tycho_rt.c) returns zero matches. The ~9 hits that motivated the phase are all substring false positives: \uXXXX -- the JSON escape notation, which contains XXX -- appears seven times in corelib/json/json.ty, once in corelib/test/json/main.ty and once in a docs/internals/FRICTION.md record line; to_XXX in tests/sized_family.ty:2 is the placeholder spelling for the conversion-function family. None is a marker. Nothing to fix; the scan is the deliverable.

Phase 2 -- LICENSE / SECURITY / CONTRIBUTING [DONE 2026-08-04 -- already present]

Phase 3 -- README: the seven tools as evidence of exercise [DONE 2026-08-04]

Added the "Eight programs that tested the language" subsection under The evidence: the seven tool programs with the axis each stressed and its ground-truth gate, plus the one finding the campaign did produce (hex literals, filed by the chess engine's castling masks).

Phase 4 -- the CI decision [DONE 2026-08-04 -- local-only]

The owner's call: no GitHub Actions. The local-only stance documented in the README and CONTRIBUTING stands unchanged; no workflow is added and no claim is rewritten. The repo is polished as-is.