Skip to content

GCP V49 ‐ Summarized

Jump to bottom
Pheonetic Coder edited this page Aug 15, 2025 · 1 revision

Status: Current release (V49).
Scope: AI-native invention protocol with gate/ledger governance, runner system, and continuous evaluation.

What’s new in V49 (highlights)

  • GitHub-safe charts: all diagrams pre-rendered as PNG/SVG (no Mermaid renderer quirks).
  • Universal Gate Decision Card (clean): preview before action; logs to DECISION_LEDGER; Auto-Mode respects user overrides.
  • Runner System: attach only what you need (Code, Physical, Theory, OT, Mobility, LifeSci, AgMRV, FinTech) with alias bundles (e.g., medical-ai ⇒ [Code, LifeSci]).
  • Evidence TTL: default refresh cadences (R1/R2/R3) with re-verification plan.
  • Packaging & Ledgers: consistent naming and export bundle; SBOM/signing notes.
  • Flow refinements: tightened checkpoint map from Worth-It Sprint → Docs/Handoff → Release → Continuous Evaluation.

Quick start (LLM-only)

  1. Upload the V49 manual and use the Gate/Checkpoint prompts.
  2. Choose Mode: Full Run (stop at each checkpoint) or Auto (only stops at critical gates or when flagged).
  3. At each ⛔ checkpoint, reply with:
    • proceed 1 • branch Phase <n> • return C#.# • end & export.

V49 Diagrams

Phase Map

gcp_v49_phase_map

Universal Gate Decision Card (Auto Mode UX)

gcp_v49_gate_ux (1)

Runner System & Alias Examples

gcp_v49_runners

Auto-seeded Templates by Runner

gcp_v49_templates_tree

Evidence TTL (R1/R2/R3)

gcp_v49_evidence_ttl

Packaging & Ledgers (snapshot)

gcp_v49_ledgers_table

Files & structure (selected)

  • Ledgers: Run • Decision • Assumption • Evidence • Benchmark • Optimization • Risk/Safety • Compliance • Contradiction • C_K • Morph_Futures • Adversary
  • Naming: RunID_Phase.Subphase_Artifact_Short.ext
  • Base repro kit: REPRO_PROTOCOL.md, REPRO_RESULTS.csv, ENV_LOCKFILE.yml
  • Export: RunID_FinalPackage.zip with manifest.

Migration notes (from v47.x → v49)

  • Replace old Mermaid charts with the PNG/SVG assets above.
  • Keep your existing gate logic; the Gate Decision Card is now a first-class artifact in Auto-Mode too.
  • Align risk lanes to R1/R2/R3 defaults and add TTL-based re-verification to release plans.

Runners (V49) — Overview & Usage

What this page is: a practical explainer for GCP’s optional Runner System.
Runners add domain-specific lanes (extra roles, artifacts, micro-gates) on top of the core protocol. GCP V49 runs perfectly without runners—attach only what you need.

See: GCP V49 for the full protocol and gate map.

TL;DR

  • Attach runners to enrich a run with domain scaffolding.
  • Aliases make it quick to call common combos (e.g., medical-ai ⇒ [Code, LifeSci]).
  • Auto-seeded templates drop in the right docs/files/ledgers for that domain.
  • Risk tiers (R1/R2/R3) still apply—runners just raise the floor for rigor where it matters.

Commands (copy/paste)

attach runner <Code|Physical|Theory|OT|Mobility|LifeSci|AgMRV|FinTech> detach runner <...> show runner status

Mode controls (unchanged) switch to Full Run | switch to Auto set auto_gate_verbosity = full|brief set auto_gate_preview = on|off

markdown Copy Edit

Runner Catalog (what each adds)

Code

  • Focus: software/ML/agents, SDK/CLI/API readiness.
  • Adds: ADRs, error taxonomy, test matrix, packaging notes.
  • Seeds (examples):
    • 01_code/ARCHITECTURE_DECISIONS/ADR-0001.md
    • 04_benchmarks/results/TEST_MATRIX.csv
    • 11_product/api/ERROR_TAXONOMY.md
  • Use when: you’re building code, agents, or APIs and want clean decisions, tests, and product surfaces.

Physical

  • Focus: mechatronics, EE/ME, embedded.
  • Adds: CAE summaries, hardware evidence hooks.
  • Seeds (examples):
    • 06_evidence/stats/CAE_SUMMARY.md
  • Use when: there’s hardware, firmware, or integration with physical systems.

Theory

  • Focus: axioms, conjectures, proofs, theoretical safety margins.
  • Adds: proof obligations, derivations, assumptions ledger emphasis.
  • Use when: you need formal reasoning or math-heavy derivations.

OT (Operational Technology)

  • Focus: industrial/utility systems (PLC/SCADA/DCS), standards like IEC 61131/62443/NERC.
  • Adds: zone/conduit models, SIF registers, safety contracts.
  • Seeds (examples):
    • OT/NETWORK_ZONES_CONDUITS.md
    • OT/SAFETY_SIF_REGISTER.md
  • Use when: your system touches plant/field equipment or critical infrastructure.

Mobility

  • Focus: automotive/aerospace/robotics, AUTOSAR, DO-178C/254-style rigor.
  • Adds: ODD definition, coverage reports, safety artifacts.
  • Seeds (examples):
    • ODD_SPEC.md
    • STRUCTURAL_COVERAGE_REPORT.md
  • Use when: vehicles, robots, or flight—anything with ODD and safety envelopes.

LifeSci

  • Focus: GxP CSV/CSA, 21 CFR Part 11, IEC 62304, ISO 14971.
  • Adds: validation traceability and regulated deliverables.
  • Seeds (examples):
    • VALIDATION_MASTER_PLAN.md
  • Use when: life sciences, medical, or regulated clinical/quality contexts.

AgMRV

  • Focus: agriculture & environmental Measurement, Reporting, Verification.
  • Adds: trial design, QA/QC hooks, field data handling.
  • Use when: field trials, soil/carbon projects, or environmental claims.

FinTech

  • Focus: PCI, SOX, AML/KYC/sanctions, SR 11-7 model risk.
  • Adds: controls mapping, model risk registers, audit trails.
  • Use when: financial services or anything touching payments/compliance.

Alias Map (human-friendly names)

Aliases ignore case and minor punctuation. Examples:

  • medical-ai ⇒ [Code, LifeSci]
  • industrial-robotics ⇒ [OT, Mobility]

How runners change the flow (practically)

  • Same checkpoints; extra context. You still pass gates C-series as usual; runners add domain-specific checks and docs at key gates (e.g., C7.Repro, safety docs before C9).
  • Evidence TTL still rules. Runners don’t change refresh cadence—use R1/R2/R3 defaults unless your domain mandates tighter windows.
  • Source of truth stays the ledgers. Runners seed files, but decisions always land in DECISION_LEDGER, assumptions in ASSUMPTION_LEDGER, etc.

FAQ

Do I need runners for every project?
No. They’re optional accelerators—attach only what’s valuable.

Can I attach multiple runners?
Yes. Combos are common (e.g., Code + LifeSci for regulated ML; OT + Mobility for industrial robotics).

Can I customize seeds?
Absolutely. Treat seeds as templates—edit freely; keep ledger naming conventions intact.

What if my domain isn’t listed?
Use the closest runner (often Code), then add a minimal custom template set under domain/ and attach via alias.

See also

Clone this wiki locally