Phase 0: Stabilize the foundation

[hyperpolymath]

Tracking issue for Phase 0 of the production path. See docs/PRODUCTION-PATH.adoc §Phase 0 for the full statement.

Theme

Close the engineering-surface fragility around the proofs. PRs were merging with red CI; the verifier job we added to catch wasmparser breaks was itself silently broken; the ROADMAP claimed things that weren't real. Phase 0 fixes that gap so all subsequent phases have a load-bearing foundation.

Deliverables

• All CI jobs on main either green or explicitly removed (PR CI triage: fix 4 of 6 persistent red jobs on main #46 is the first round)
  • Cargo build + test (PR CI triage: fix 4 of 6 persistent red jobs on main #46)
  • Structural E2E (PR CI triage: fix 4 of 6 persistent red jobs on main #46)
  • Build + E2E (Idris2 + Zig) — Zig URL fix landed; still needs runtime verification
  • governance / Language / package anti-pattern policy — blocked on Track A's ReScript removal
  • Validate A2ML manifests — third-party action, needs investigation
  • Validate K9 contracts — third-party action, needs investigation
• Track A — Codegen pipeline kicked off
  • tree-sitter-twasm grammar derived from spec/grammar.ebnf, in-tree at tools/tree-sitter-twasm/
  • Idris2 parser at parity with ReScript parser (188-test corpus)
  • ReScript cut (single PR removing all .res + rescript.json + npm deps + workflow steps)
  • Codegen v0 for examples/01-single-module.twasm end-to-end
• Track B — AffineScript verifier migration
  • Cross-repo PR swapping affinescript/lib/{tw_verify,tw_interface}.ml for subprocess call to typed-wasm-verify binary
  • Retire ~491 LOC of OCaml verifier logic
• Track C — Audit-floor cleanup
  • cargo audit job on GitHub CI
  • Build real tests/property/property_test.mjs (the TEST-NEEDS ghost)
  • Add Security aspect-test dimension to tests/aspect/
  • Add proof-level regression tests so an Idris2 toolchain bump can't silently break a theorem
• ROADMAP truthfulness audit: every "DONE" claim verified against a green CI run

Gate to Phase 1

1. Every commit on main exits CI green (or the gate is explicitly removed from the merge requirement set, with reason recorded in PRODUCTION-PATH.adoc)
2. Codegen v0 emits valid wasm for examples/01-single-module.twasm, verifiable end-to-end by typed-wasm-verify
3. ROADMAP.adoc reflects reality

In-flight PRs

• CI triage: fix 4 of 6 persistent red jobs on main #46 — CI triage (3 of 6 red jobs fixed)
• docs: 6-phase production path + version↔phase mapping #47 — this plan documented in the repo

Status when this issue closes

The project is defensibly engineered — still a research prototype but one that builds, tests, and ships without hand-holding. Not yet a working compile target; that's Phase 1.

[hyperpolymath]

Phase tracking map

The full 6-phase production path is documented in docs/PRODUCTION-PATH.adoc (landing via #47). Tracking issues:

Phase	Issue	Theme	Gate to next
0	#48 (this)	Stabilize foundation	CI green; codegen v0 works; ROADMAP truthful
1	#49	End-to-end producer	.twasm → .wasm round-trips for all examples
2	#50	Multi-producer adoption	≥3 producers shipping verified wasm
3	#51	Runtime-side enforcement	Reference runtime detects L7+ violation verifier missed
4	#52	Tooling + DX	Outside-ecosystem ship without maintainer support
5	#53	Spec + standards	1.0 frozen; conformance suite; academic publication
6	#54	Production hardening	SLA + CVE + ≥1 production deployment + case study

Suggest creating a GitHub Project board grouping by the phase:N labels — that's the canonical place to see what's in flight per phase. (GitHub MCP doesn't expose project-board creation directly; the labels are in place to make the board straightforward to set up from the UI.)

In-flight Phase 0 PRs

• CI triage: fix 4 of 6 persistent red jobs on main #46 (merged) — CI triage round 1: cargo, structural, hypatia exemption
• docs: 6-phase production path + version↔phase mapping #47 — adds docs/PRODUCTION-PATH.adoc as the canonical plan + maps version axis to phase axis

Six load-bearing decisions

PRODUCTION-PATH.adoc §"Load-bearing decisions" lists six choices due in the first 12 months that determine which phases are reachable. Each should land an ADR under docs/decisions/ when made:

1. Stay codegen-on-top-of-wasm vs. own bytecode
2. Producer-side-only vs. runtime-aware
3. W3C CG path vs. independent ecosystem
4. Idris2-only proofs vs. dual implementation
5. MPL-2.0 vs. dual MPL/Apache
6. Single-maintainer vs. recruit committers

---

Generated by Claude Code