Sprint 5 closes cleanly at ≥ 0.85. Between 0.70–0.84 it can conditional-close with explicit Ken ratification on named exceptions. Below 0.70 it stays open with a remediation plan. Right now: the two biggest movers are the 9 required tasks (9/9 done) and authoring the Sprint 6 candidates source (-30 days to deadline).
Per Section 1 of state/projects/sprint-5-close-criteria.md.
Each must be in project_tasks.status='done' (or 'shipped') by close.
The required-tasks component carries 35% of the rubric weight — by far the largest single mover.
What it is · Rewrite the Memex extraction watchdog so it tracks an accumulating counter rather than a snapshot. Without this, the watchdog can lose count when restarted mid-job and report misleading 'progress'.
Why in Sprint 5 · Memex Phase 1 close criterion — the watchdog has to survive crashes cleanly to declare the chain green.
What it is · Tune token budget per LLM turn for Memex's summarization step so it doesn't burn excess context on routine chunks. Reduces overall Anthropic burn for Memex ingestion runs.
Why in Sprint 5 · Cost discipline — keeps Memex Phase 1 affordable as the corpus grows.
What it is · Build a manual-sync v0 that parses `state/calendar/work-events.md` (Ken-edited markdown) into the calendar_events table — fallback when ICS sync isn't available (e.g., when Ken's at a hotel).
Why in Sprint 5 · Calendar travel-awareness — Ken needs the briefing to work even off-network.
What it is · Skeleton route + spec file for the future `/evening` CP page (sister to `/morning`). The route doesn't ship rich functionality yet — it's a placeholder so subsequent work has a target. (Note: `/evening` evening-prep copilot button shipped tonight separately as a partial fulfillment.)
Why in Sprint 5 · Architecture parity — every recurring ritual surface needs an explicit route, not just morning.
Blocked on · Phase 1 Claude checklist usage data — 5-7 evening cycles required
What it is · Make the morning-briefing → Reflect-daily-note insertion idempotent via Reflect's MCP edit tools. Re-running the morning pipeline can't post duplicate content.
Why in Sprint 5 · Critical Invariant — the morning briefing posts every day; double-posts are an unacceptable Ken-frustration.
What it is · Author Ken's Mayo Clinic prep summary doc — health context Ken hands to his Mayo PCP at the May 29 visit. Pulls from biography health timeline + recent labs.
Why in Sprint 5 · External deadline (Mayo visit) sits just past Sprint 5 close — the doc needs to be drafted in this sprint to allow review/refinement before May 29.
What it is · Extract small OneNote notebooks (Jacynda + Reference + small Personal OneDrive) to .md files via the OneNote-to-md export tool. Quick-win compared to the larger 522-page extraction.
Why in Sprint 5 · Memex Phase 1 chain checkpoint #2 — needed for OneNote source coverage.
What it is · Extract Kevin Lovell Mission OneNote notebook (~100 pages) to .md. Personal/family-history content that feeds biography PKM.
Why in Sprint 5 · Memex Phase 1 chain — the ~640 remaining OneNote pages get in via three tasks (153/154/155); each one closes a chain checkpoint.
What it is · Extract Ken's Personal OneDrive notebook (~522 pages, batched). Largest of the three OneNote extraction tasks — the bulk of the bio-source corpus.
Why in Sprint 5 · Memex Phase 1 — the load-bearing OneNote extract. Without this, biography PKM has a major coverage gap.
Confidence = Σ(weight × score). Each component scores 0.00–1.00. Click any component below for the full plain-language explainer + action steps.
What it means · 9 named must-ship tasks for Sprint 5 close (per Section 1 of sprint-5-close-criteria.md). Score is fraction with status 'done' or 'shipped' in project_tasks. This is the largest rubric weight at 35% — by far the biggest mover of the headline confidence.
Why now · Sprint 5 close target 2026-05-10. With this at 0.00 and 35% weight, the headline cannot exceed 0.65 even if every other component is perfect. Ship one required task → headline jumps ~0.04. Three → ~0.12.
Current evidence · 6 of 6 required tasks shipped/done
What it means · 7 checkpoints in the Memex Phase 1 chain (per Section 3 of sprint-5-close-criteria.md): watchdog verified, OneNote small notebooks, OneNote 522-page Personal, OneNote bio source, R5-1 idempotent insert, smoke query green, closure dispatch. Score is fraction green.
Why now · Memex was declared LIVE 2026-04-30 10:40 ET. The remaining non-green checkpoints are mostly OneNote-blocked (3 OneNote extraction tasks). 5/7 holds the score at ~0.71; getting all three OneNote tasks done lifts it to 1.00 (full 20% credit).
Current evidence · 7 of 7 Memex Phase 1 checkpoints green
What it means · 5 named Ken-decisions required before Sprint 5 close (per Section 4 of sprint-5-close-criteria.md). Score is fraction resolved. v1 proxies the count via pm_decisions where decided_at IS NULL — score=1.00 means 0 pending.
Why now · Each unratified Ken-decision blocks downstream mini work. If the proxy says 1.00 today, no specific decisions are currently pending in pm_decisions. New decisions get logged as they're proposed — this score will drop when one lands.
Current evidence · 5 of 5 Ken-decisions resolved (0 pending in pm_decisions)
What it means · BINARY component — does a `Documents/Sprint 6 Candidates*.md` file exist by 2026-05-05? Score is 1.00 if yes, 0.00 if no. 10% rubric weight. The single biggest single-action lift available: one file's existence buys 10 percentage points.
Why now · Hard deadline 2026-05-05 (5 days out). Without this file by then, the slip_clean component (also 5%) ALSO stays at 0.00 since it's binary-tied to this file existing.
Current evidence · authored: sprint-6-plan-FINAL-LOCKED-2026-05-02.md
What it means · Auto-correction sweep state — total findings + open RULE-9 (ambiguous) findings. Score is 1.0 - (0.1 × open_rule_9_count); small penalty per open ambiguous finding. 10% rubric weight.
Why now · Currently green at 1.00 — sweep is healthy with 0 open RULE-9. Component drops by 0.10 per new open ambiguous finding. Watch this on each cadence run.
Current evidence · 125 sweep findings, 0 open RULE-9
What it means · Recent SHIPPED dispatches must include E2E evidence (per RULE 11 + RULE 12). Score is fraction of last-7-days mini-to-* dispatches claiming 'shipped'/'live'/'complete' that ALSO contain at least one of: curl, http, launchctl, sqlite3, smoke, evidence, exit code, RULE 12. 5% rubric weight.
Why now · Discipline trends compound — ships without evidence are claims, not facts. A ratchet down here is how Sprint 5 'closes' on claims rather than ground truth.
Current evidence · 70 of 192 ships have evidence
What it means · BINARY — same gating signal as sprint_6_candidates. Score is 1.00 if a `Documents/Sprint 6 Candidates*.md` file exists, else 0.00. 5% weight. The premise: items that won't ship in Sprint 5 must land in the Sprint 6 source so they're not lost.
Why now · Tied to the same 2026-05-05 deadline as sprint_6_candidates — fixing one fixes both. Combined they're 15% of the rubric. 40 currently-slipping tasks are auto-included in the draft the helper generates.
Current evidence · slip-cleanup in place
JSON: /api/pm/sprint/5/json · Live cadence: /api/pm/sprint-close · Reflect note: CP — Live Status ↗