docs: v0.4.1 positioning — Evidence as Code frame

[avrabe]

Summary

• New file docs/what-is-rivet.md — the canonical positioning doc (~2000 words). Frames rivet as the audit substrate for AI-assisted engineering ("Evidence as Code"), contrasts with Sphinx-Needs' "Engineering as Code", and gives a per-situation playbook for 11 use-cases (TDD, ASPICE, STPA/ISO 26262, requirements, variant/PLE, LLM code review, provenance, cross-tool interop, GSN, tool qualification, spec-driven dev). Each use-case is ~100 words with question / artifacts / AI role / human role / explicit limits.
• Rewrite top of README.md — replaces the "SDLC traceability for safety-critical systems" + feature-list first paragraph with an evidence-centric pitch that points at docs/what-is-rivet.md for depth. The rest of the README (features, quick start, CLI table, architecture, dogfooding, dev) is untouched.
• Explicit human vs AI split table — authoring, linking, validation, coverage, variant, provenance, compliance export, tool qualification, schemas.
• Explicit limits section — no Polarion live editing, no direct ALM connector today, no iso-26262.yaml schema yet, no npm distribution yet. All in-flight items are marked "planned for v0.5.0".

All capability claims are verifiable against main HEAD:

• 34 hazards + 62 UCAs + 62 controller constraints (safety/stpa/)
• MCP tool set matches rivet-cli/src/mcp.rs (rivet_list/get/query/add/modify/link/unlink/remove/validate/coverage/stats/schema/embed/snapshot_capture/reload)
• s-expr predicates forall / exists / reachable-from / reachable-to match rivet-core/src/sexpr_eval.rs
• 27 Kani harnesses in rivet-core/src/proofs.rs
• 324 Playwright tests across 28 spec files in tests/playwright/
• Schemas referenced (iso-pas-8800, iec-61508, iec-62304, do-178c, en-50128, eu-ai-act, safety-case, aspice, stpa, stpa-sec) all exist in schemas/

Chosen one-sentence pitch: "rivet is the audit substrate for AI-assisted engineering — every artifact, link, and decision carries evidence a human can review in a pull request."

Alternates considered:

• "rivet lets AI agents build your traceability chain; humans review the PR."
• "Evidence as Code: rivet turns LLM-authored SDLC work into a git-native, schema-validated audit trail."

Trailer: Refs: FEAT-001.

Test plan

• rivet validate still PASS (docs changes only; no schema/source changes)
• Links inside docs/what-is-rivet.md resolve — especially design/polarion-reqif-fidelity.md, getting-started.md, architecture.md, schemas.md, stpa-sec.md, verification.md, roadmap.md, audit-report.md
• README.md renders cleanly on GitHub (no broken reference badge or link)
• The "planned for v0.5.0" markers are consistent between README, what-is-rivet, and the roadmap

🤖 Generated with Claude Code

[github-actions]

⚠️ Performance Alert ⚠️

Possible performance regression was detected for benchmark 'Rivet Criterion Benchmarks'.
Benchmark result of this commit is worse than the previous benchmark result exceeding threshold 1.20.

Benchmark suite	Current: 7812ece	Previous: 61bfc41	Ratio
store_lookup/100	2243 ns/iter (± 8)	1677 ns/iter (± 3)	1.34
store_lookup/1000	24913 ns/iter (± 182)	19262 ns/iter (± 54)	1.29
traceability_matrix/1000	60215 ns/iter (± 161)	40570 ns/iter (± 154)	1.48
query/100	820 ns/iter (± 2)	646 ns/iter (± 2)	1.27
query/1000	7511 ns/iter (± 28)	5450 ns/iter (± 12)	1.38

This comment was automatically generated by workflow using github-action-benchmark.