Contents
- pgmnemo — Tactical Plan, Month 1
- Sprint goal of Month 1
- Week 1 (T0 → T+1w): Charter + License + Plans
- Week 2 (T+1w → T+2w): Schema + Migration draft + Bench harness
- Week 3 (T+2w → T+3w): Retrieval functions + Provenance gate + Demo case 1
- Week 4 (T+3w → T+4w): Demo case 2 + Internal review + Phase 1 measurement
- Month 1 budget total
- Risks (Month 1 specific)
- Outputs at end of Month 1
- Working Group involvement schedule
pgmnemo — Tactical Plan, Month 1
Window: T0 (2026-04-29) → T+4w (2026-05-27) | Owner: technical_lead (5) with PO sign-off
Sprint goal of Month 1
By end of Week 4, have: license locked, repo bootstrapped (private), Phase 1 design complete, schema migrations drafted, retrieval functions implemented and dogfooded internally on Agency v2, positioning v1.0 ready for friendly review.
Month 1 is internal-only. No public launch this month. Public is Month 2.
Week 1 (T0 → T+1w): Charter + License + Plans
Goal: all foundational artifacts exist; license locked; Phase 1 plan reviewed and approved.
| # | Task ID | Title | Owner | Cost | Due |
|---|---|---|---|---|---|
| W1.1 | 2114 | LICENSE_DECISION_LOCKIN | legal_advocate (74) | $1.00 | T+3d |
| W1.2 | 2115 | BUILD_MVP_EXT_PHASE1_PLAN | technical_lead (5) | $3.00 | T+4d |
| W1.3 | 2116 | LAUNCH_POSITIONING_v0.1 | growth_lead (92) | $2.00 | T+4d |
| W1.4 | 2117 | MENTOR_REVIEW_T0 | startup_mentor (91) | $2.00 | T+3d |
| W1.5 | 2118 | PRODUCT_PLAN | product_owner (16) | $3.00 | T+4d |
Gate at end of Week 1: founder reviews POSITIONING + PRODUCT_PLAN + MENTOR_REVIEW. Go/hold decision.
Week 1 budget: $11.00 cumulative.
Week 2 (T+1w → T+2w): Schema + Migration draft + Bench harness
Goal: extension scaffolding exists in private repo; schema migrations applied; benchmark harness ready to compare against OpenBrain.
| # | Task | Owner | Cost (est) |
|---|---|---|---|
| W2.1 | Extension scaffolding: pgmnemo.control, pgmnemo--0.0.1.sql, Makefile |
backend_developer (70) | $2.00 |
| W2.2 | Schema design: agent_lesson, memory_concept, provenance_log tables + indexes |
chief_architect (86) | $3.00 |
| W2.3 | Migration v1.0: schema + base seed data | backend_developer (70) | $2.00 |
| W2.4 | Benchmark harness setup: BL-B + OpenBrain comparison fixtures | experiment_designer (84) | $3.00 |
| W2.5 | Repo bootstrap (private) — founder runs REPO_BOOTSTRAP_CHECKLIST | founder | $0 |
Week 2 budget: $10.00. Critical path: W2.5 (founder action) blocks public-facing W2.½.2 commits — but they happen in private repo from T+1w onward, so blockers are minimal.
Week 3 (T+2w → T+3w): Retrieval functions + Provenance gate + Demo case 1
Goal: core retrieval works; provenance gate trigger fires; first end-to-end demo runs.
| # | Task | Owner | Cost (est) |
|---|---|---|---|
| W3.1 | recall_lessons() PL/pgSQL function with tsvector + trigram |
backend_developer (70) | $3.00 |
| W3.2 | search_concepts() PL/pgSQL function with role/project filtering |
backend_developer (70) | $2.00 |
| W3.3 | Provenance gate trigger — verify commit SHA / artifact hash before promotion | backend_developer (70) | $4.00 |
| W3.4 | Demo case 1: Claude Code agent reading from pgmnemo (script + README) | technical_lead (5) | $2.00 |
| W3.5 | PAPER v0.3 outline aligned with measured Phase 1 results | principal_investigator (77) | $3.00 |
| W3.6 | MENTOR_REVIEW T+2w (biweekly cadence) | startup_mentor (91) | $2.00 |
Week 3 budget: $16.00. Risk: W3.3 (provenance gate) is the technical USP — if it slips, all of Month 2 timeline slips.
Week 4 (T+3w → T+4w): Demo case 2 + Internal review + Phase 1 measurement
Goal: Phase 1 acceptance gates measured; second demo (multi-agent role routing) shipped internally; team alignment for Month 2 public launch.
| # | Task | Owner | Cost (est) |
|---|---|---|---|
| W4.1 | Demo case 2: multi-agent role routing across 3 agents | technical_lead (5) + backend_developer (70) | $3.00 |
| W4.2 | Phase 1 measurement: recall@10, install time, memory footprint, query p95 | experiment_designer (84) + statistical_analyst (79) | $2.00 |
| W4.3 | Phase 1 measurement vs OpenBrain (head-to-head benchmark) | experiment_designer (84) | $2.00 |
| W4.4 | Internal QA pass — fresh PG container install + smoke test | qa_set (6) | $1.00 |
| W4.5 | PAPER v0.3 first draft (submission-ready) | paper_writer (81) | $4.00 |
| W4.6 | Month-1 retrospective + Month-2 Go/Stop decision (WG vote) | technical_lead (5) | $1.00 |
Week 4 budget: $13.00. Decision gate: Founder + WG vote at end of Week 4 — proceed to Month 2 public launch yes/no?
Month 1 budget total
| Week | Budget | Cumulative |
|---|---|---|
| W1 | $11.00 | $11.00 |
| W2 | $10.00 | $21.00 |
| W3 | $16.00 | $37.00 |
| W4 | $13.00 | $50.00 |
Month 1 cap: $60 (20% buffer over $50 baseline). If burn exceeds $60 before W4 retrospective → mandatory founder ack to continue.
Risks (Month 1 specific)
| # | Risk | Owner | Mitigation |
|---|---|---|---|
| R1 | Provenance gate (W3.3) more complex than estimated | TL | If 2x overrun → defer to W4, reduce demo scope |
| R2 | OpenBrain benchmark hard to set up (W2.4) | ExpDesigner | Fall back to internal-only benchmark; flag for Month 2 |
| R3 | Founder unavailable for repo bootstrap (W2.5) | Founder | Pre-create bootstrap-files/ so action is 5 minutes |
| R4 | License decision drags > Week 1 | legal_advocate | Default to Apache-2.0 + DCO sign-off if no objection by T+5d |
| R5 | Two-product split (Agency + pgmnemo) creates context-switch overhead | TL + PO | Use Agency v2 as pilot user — same codebase, single track |
Outputs at end of Month 1
By 2026-05-27 the following must exist in spec/v2/pgmnemo/:
spec/v2/pgmnemo/
├── STRATEGY.md ✓ (created T0)
├── TACTICAL_M1.md ✓ (this file, created T0)
├── PRODUCT_PLAN.md (PO deliverable W1)
├── LICENSE_DECISION.md (legal deliverable W1)
├── POSITIONING.md (growth deliverable W1)
├── REPO_BOOTSTRAP_CHECKLIST.md ✓ (created T0)
├── MENTOR_REVIEW_2026-04-29.md (W1)
├── MENTOR_REVIEW_2026-05-13.md (W3)
├── MENTOR_REVIEW_2026-05-27.md (W4 → Go/Stop)
├── design/
│ └── BUILD_MVP_EXT_PHASE1.md (TL deliverable W1)
├── research/
│ └── (7 frozen copies from memory-svc/, see REPO_BOOTSTRAP_CHECKLIST)
└── (private repo) extension code in github.com/pgmnemo/pgmnemo
Working Group involvement schedule
| Member | W1 | W2 | W3 | W4 |
|---|---|---|---|---|
| product_owner (16) | ★ | — | ★ | ★ |
| technical_lead (5) | ★ | ★ | ★ | ★ |
| chief_architect (86) | — | ★ | — | ★ |
| backend_developer (70) | — | ★ | ★ | ★ |
| principal_investigator (77) | — | — | ★ | — |
| paper_writer (81) | — | — | — | ★ |
| literature_scout (82) | — | — | — | — |
| experiment_designer (84) | — | ★ | — | ★ |
| statistical_analyst (79) | — | — | — | ★ |
| growth_lead (92) | ★ | — | — | — |
| legal_advocate (74) | ★ | — | — | — |
| qa_set (6) | — | — | — | ★ |
| startup_mentor (91) | ★ | — | ★ | ★ |
★ = active week. Average burn-rate sanity: ~5 active people × 10 hours/week × $5 task budget = $250/week capacity, far above $13/week task budget — so concurrency is not the bottleneck.