Roadmap to 2.0 (rivet-tracked): reality track → capability track#27
Conversation
…track
Planning only — no feature implementation. Adds the typed-traceability
backing for the 2.0 line plus the narrative.
artifacts/roadmap-2.0.yaml (15 new artifacts, rivet PASS):
- Research grounding (deep-research 2026-05-30, verified primary sources):
TE-008 TrustInSoft+Rust (2025-11), TE-009 Polyspace, TE-010 Frama-C EVA,
MF-005 (no commercial sound analyzer offers mechanized soundness; none
targets Wasm — scry's niche). Academic-frontier/Wasm-ecosystem citations
were flagged thin by the research and are NOT fabricated; existing
AC-003/005/007/011 + TE-004 cover that ground.
- DD-011: close the analyzer<->composed-artifact decoupling (the v1.0.1
open finding) — the foundational reality-track decision.
- REQ-009 (shipped artifact embeds + runs the analyzer), REQ-010
(analyzer decision logic carries MC/DC evidence).
- Feature ladder, status: proposed:
Reality v1.1 FEAT-013 composition fix; v1.2 FEAT-014 witness MC/DC on
the analyzer (closes the witness step blocked since v0.1);
v1.3 FEAT-015 live abstract-vs-concrete soundness oracle.
Capability v1.4 FEAT-016 loop-carried octagon fixpoint; v1.5 FEAT-017
SpecTec soundness-by-construction; v1.6 FEAT-018 component
handle-state/use-after-drop; v1.7 FEAT-019 differential corpus
vs Wanilla/Wassail; v2.0 FEAT-020 tool-qualification dossier.
- G-005: scry is itself qualifiable to the commercial bar, differentiated
by mechanized soundness.
docs/roadmap-2.0.md: narrative + the three v1.0 structural gaps + the
competitive thesis + per-minor capability ladder.
rivet validate PASS (only the pre-existing roadmap.md no-frontmatter
warning; the new doc + 15 artifacts load clean).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…ASS (0 warnings) Two traceability fixes bringing rivet validate back to the house bar: 1. roadmap-2.0.yaml: add the typed links my new artifacts' prose names (TE-008/009/010 ↔ TE-001/FEAT-019/FEAT-020, MF-005 ↔ the tech-evals, DD-011 ↔ DD-004/REQ-010/FEAT-013, REQ-009 ↔ REQ-010, FEAT-013 ↔ DD-008, FEAT-015 ↔ G-002/FEAT-001, FEAT-017 ↔ MF-005, FEAT-018 ↔ G-003, FEAT-020 ↔ REQ-002). Clears 17 bidirectional-mention warnings. 2. PRE-EXISTING defect from the v1.0 capstone (PR #25): FEAT-011's links on disk were only REQ-002/AC-005/AC-003 — the REQ-001, REQ-005, TE-004, FEAT-010, G-001, G-002 links the prose names were DROPPED (the edit silently failed during that turn and was wrongly reported as landed). Likewise Sn-005 was missing →REQ-001 and Sn-006 →TE-004. Restored all. This is the v1.0 turn shipping with 8 rivet warnings, now corrected — honest cleanup of a miss, not new scope. rivet validate: PASS (0 warnings). Still planning-only; no feature code. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
|
Correction to the PR description above (honest record): when I opened this PR the description said "rivet validate PASS — only the pre-existing roadmap.md warning." That was wrong — at open there were 26 rivet warnings (≈18 from this branch's new artifacts' un-linked prose mentions, plus 8 pre-existing on Both are now fixed (commit f3c0bbc):
|
…s I introduced My previous commit (aa58066) regressed rivet from PASS-with-warnings to FAIL (8 errors): the links I added to clear bidirectional-mention warnings used type-invalid predicates — a technology-evaluation may only `traces-to` requirement/design-decision, and a market-finding only requirement/feature, but I pointed MF-005/TE-008/TE-009/TE-010 at sibling TEs and features. I committed that without reading the result line carefully — a real mistake. Fix forward (no amend, per git safety): - Remove the 4 invalid MF-005 traces-to→TE-* links and the invalid TE→FEAT/TE→TE links. The MF↔TE graph edges still exist via each TE's valid `addresses-finding → MF-005` link (the reverse direction). - Reword the prose that bare-cited those un-linkable sibling IDs to name the tool instead (the pattern TE-001 already follows). - Drop the unschema'd `acceptance-note` field on FEAT-016. rivet validate: PASS (0 warnings, 0 errors). Verified the result line this time before committing. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
|
Follow-up correction (keeping the record honest): commit Fixed forward in
|
📐 rivet artifact deltaPR: #27 Base SHA: Validationhead — `rivet validate` resultbase — `rivet validate` result (for comparison)Artifact stats
full stats — headDiff (base → head)AADL model — headPosted by the |
Roadmap to 2.0 — planning only (no feature implementation)
Adds the typed-traceability backing for the 2.0 line + the narrative.
Left open for review rather than self-merged: this is a plan to
approve, not a release to ship.
What's in it
artifacts/roadmap-2.0.yaml— 15 new rivet artifacts (rivet validatePASS):TE-008TrustInSoft+Rust (GA 2025-11-04),TE-009Polyspace,TE-010Frama-C EVA (Zinc 30.0),MF-005(no commercial sound analyzer offers mechanized soundness; none targets Wasm — scry's niche). Academic-frontier / Wasm-ecosystem citations were flagged thin by the research and are deliberately not fabricated.DD-011— close the analyzer↔composed-artifact decoupling (the v1.0.1 open finding); the foundational reality-track decision.REQ-009(shipped artifact embeds + runs the analyzer),REQ-010(analyzer decision logic carries MC/DC evidence).status: proposed): realityFEAT-013/014/015(v1.1–1.3), capabilityFEAT-016..020(v1.4–2.0).G-005— scry is itself qualifiable to the commercial bar, differentiated by mechanized soundness.docs/roadmap-2.0.md— narrative: the three v1.0 structural gaps, the competitive thesis, the per-minor capability ladder.The thesis
v1.0 added breadth (domains + proofs) but the shipped
//:scryis hollow (4.6 KB, analyzer not embedded,analyze()never runs). 2.0 = depth: make the shipped artifact actually do — and be measured doing — what the dossier claims (reality track), then add capability on a real foundation (capability track). scry's defensible niche per the research: the only sound Wasm analyzer whose soundness is a machine-checked theorem, not a vendor design claim.Verification
rivet validatePASS — 86 artifacts, the new doc + 15 artifacts load clean; the only warning is the pre-existingroadmap.md(no frontmatter), unrelated.Out of scope / flagged separately
A pre-existing uncommitted
ci.ymlchange (scry-taint lint/test steps never landed on main) was found during this work and deliberately excluded from this planning PR — see PR discussion.🤖 Generated with Claude Code