JIT-4 The JIT-4 platform

Web5 — What if we adopt MM2?

An AI-guided implementation of MM1 for SELLING/public sites—under the principal site’s guidance

... to make SaaS composable and truth portable ... fast.

Radical proviso (scope & aim)

Scope: Public-facing, SELLING websites.

Aim: Make the open web commercially usable quickly—without redesigns—by combining a SaaS-first spine with a thin portability layer for truth.

One-line statement

MM2 = “AI-guided MM1.” An AI implementer produces a correct MM1 layer—exact-line landing, verification, drift signal, locale-true mapping, and a tiny changefeed—under the principal site’s guidance, so a selling site can operate with composable SaaS in ≈15 days (as a reference window).

Normative requirements (strengthening Web 5)

  • RP1 — SaaS-first. Commerce capabilities (auth, catalog, checkout, payments, tax, search) MUST be swap-able services.
  • RP2 — Portability of truth. Buyer-relevant claims MUST be addressable at the sentence level and verifiable after edits/migrations.
  • RP3 — MM2 availability. For SELLING sites, an AI implementer SHOULD be able to generate the MM1 layer within a reference 15-day window, with human approval gates.
  • RP4 — Governed automation. Sites MUST publish a charter.json (consent, limits, locales, privacy) that automation MUST honor.
  • RP5 — Reversibility. Changes MUST be exportable/rollbackable; links/proofs MUST remain valid across vendor/theme/platform changes.

What MM2 actually does (under guidance)

  • Discovers visible text; proposes the 10–50 decision-driving claim lines per page.
  • Assigns stable paragraph IDs; computes sentence short-hashes; emits TAP fragments (#p;s;h).
  • Builds locale-true maps across languages; flags gaps for review.
  • Publishes a lightweight prooffeed.json (added/updated/removed + timestamps).
  • Packages runtime hints (open tab/accordion, scroll, highlight) for deterministic landing.
  • Generates exact-line links for support/search and an exec-forwardable Proof Page.
  • Preps a minimal SaaS spine config (payments/tax/search) without lock-in.

Human remains in charge: propose → review → stage → ship (approval gates at each step).

Happy-path sequence (ASCII)

Principal Site (charter.json, sitemap)
    └─► AI Implementer (MM2)
          ├─ propose claim lines (10–50/page)
          ├─ assign Line-IDs + TAP (#p;s;h) + hashes
          ├─ map locales (line-true); flag gaps
          ├─ emit prooffeed.json (diffs)
          ├─ generate UX hints (open, scroll, highlight)
          ├─ draft support/search links; Proof Page
          └─ suggest SaaS spine (swap-able)
                ▼ (approval gates)
Controller / Runtime
          ├─ verify hash; apply hints; drift badge
          └─ log time-to-clarity & verified-link events
                ▼
Users / Agents / SaaS land on the exact line, anywhere

What can be injected (extensibility)

  • Verification: semantic-drift classifier; severity badges.
  • Locale: glossary alignment; deterministic fallbacks (clearly labeled).
  • SaaS edges: payments failover; regional tax; swap search vendors without breaking links.
  • Governance: rate limits; bot identity; audit logs; robots/CSP/Permissions-Policy compliance.
  • Accessibility: focus/ARIA/live-region; reduced-motion; high-contrast highlight.

Why people would want this

  • Users: one click to the sentence that matters, in their language, with visible recency.
  • Tiny teams: no redesign; automate the boring; approve the risky.
  • Executives: forward proofs, not PDFs; measurable truth.
  • Vendors: safer trials (no hostage UX); easier integrations (stable claim anchors).
  • Regulators: provenance and change visibility at the claim level.

Acceptance criteria (MM2 over MM1)

  • Verified-link rate ≥ 95% after edits/migrations.
  • Time-to-clarity ≤ 3s to open correct UI state + highlight.
  • Coverage ≥ 85% of top questions mapped to exact lines.
  • Locale parity ≥ 95% for mapped lines (label gaps).
  • Drift detection ≤ 15 min to visible badge/feed.
  • Automation precision ≥ 90% on line selection; human approvals < 2h/day during the 15-day reference window.
  • Zero hostage UX: vendor/theme swaps preserve line links & proofs.

Safety & governance

  • Human-in-the-loop: staged approvals; reversible deploys.
  • Privacy: no PII scraping; honor robots and charter limits.
  • Security: CSP-friendly; no fingerprinting; deterministic fallbacks.
  • Internationalization: line-true first; labeled fallbacks; glossary control.
  • Reversibility: full export/rollback; signed change/audit logs.

Browser path

Today: ships as a tiny JS/edge layer (progressive enhancement). Tomorrow: candidates for native adoption—content-addressed fragments with short-hash verify, standard open-and-highlight hints, sentence-level changefeed, and locale-true resolution.

Tagline: Browser-ready today · Browser-native tomorrow.