# Pass 8 — Entropy indicators

Date: 2026-05-16

## Per-task entropy flags

Counted across all 110 pending tasks:

| Indicator | Count | Notes |
|---|---|---|
| duplicate_candidate | 3 | shared-user-dropdown, multi-trader-scorecards, phase-4-trader-selector (cluster: trader-picker overlap) |
| conflicting_architecture | 9 | 5 risk-hud-* + 4 acct-v2-hud-* (all gated on architectural decisions) |
| stale_wording | 3 | catalog-refactor-shared-json (title path), restore-0416, rescue-localstorage |
| repeated_request | 2 | p0-identity-fn (created 4/17, still open 29d); mgp-recap-stability (superseded by newer task on same theme) |
| unclear_ownership | 7 | mgp-cron-skipped-5-14 (owner casing) + 5 property mis-assignments + 1 priority-missing (multi-trader-scorecards) |
| stale_no_note_update | ~50 | All tasks ≥14d with no STATUS bump in note. Reflects normal "create + work later" pattern. Not actionable individually; per-property bumping below. |

## Per-property entropy metrics

Built from prop_counts + completion_velocity_30d + open-age stats:

| property | open count | completed_30d | velocity | oldest open (days) | cross-property overlap |
|---|---|---|---|---|---|
| **scorecards** | 27 | 18 | 0.40 | 28 (journal-user-steps) | HIGH (5 tasks live here that belong to strategy_library or mgp per drift.md) |
| accountability_hud | 14 | 6 | 0.30 | 9 (acct-v2 cluster) | LOW (well-scoped to live cockpit) |
| foundation | 13 | 7 | 0.35 | 29 (p0-* deliverables) | MEDIUM (some risk_hud tasks tagged foundation when they should be risk_hud) |
| mgp | 15 | 29 | 0.66 | 27 (mgp-multi-tf-bias) | MEDIUM (3 mgp-strategy-templates tasks tagged scorecards) |
| strategy_library | 11 | 4 | 0.27 | 28 (strategy-library-ddl) | HIGH (3 strategy-* tasks tagged scorecards) |
| playbooks | 8 | 10 | 0.56 | 28 (chart QA + invariants) | LOW |
| pm_system | 7 | 7 | 0.50 | 27 (pm-nightly-task) | LOW |
| risk_hud | 6 | 0 | 0.00 | 25 (risk-hud-stop-gauge) | HIGH (conflicting_architecture with accountability_hud) |
| param_surface | 3 | 1 | 0.25 | 29 (p0-param-surface-registry) | LOW |
| live_classifier | 2 | 2 | 0.50 | 10 (roberto tasks) | LOW |
| daily_stop | 1 | 3 | 0.75 | 27 (mgp-daily-stop-pipe) | LOW |
| mtf_vp_scanner | 1 | 1 | 0.50 | 10 (path-b handler) | LOW |
| monthly_reports | 1 | 3 | 0.75 | 10 (per-trader view) | LOW |
| alerts-hotkeys-etc | 1 | 0 | 0.00 | 4 (hot-key-managment) | inactive |

**Velocity** = completed_30d / (completed_30d + open). High velocity (>0.5) means the property is actively shipping work. Low velocity (<0.3) means either the work is genuinely hard, the property is overloaded, or both.

### Property-level entropy verdict

- **scorecards: HIGH ENTROPY.** 27 open, velocity 0.40, the highest task count + many cross-property tasks living here. Re-homing 5 tasks per drift.md + closing webhook-cluster gating drops this to ~16 open + raises velocity. **Top consolidation candidate.**
- **risk_hud: HIGH ENTROPY (different kind).** 0 completed in 30 days + 5 build tasks blocked on a decision + conflicting_architecture with the active acct-v2 HUD. The dedup decision is the leverage move.
- **strategy_library: MEDIUM ENTROPY.** 11 open, velocity 0.27 (lowest among active props). Most pending tasks gate on add-user-role-approval + catalog-refactor. Will accelerate after foundation lands.
- **foundation: MEDIUM ENTROPY.** Oldest pending (p0-* tasks 29 days). Velocity 0.35 reflects that the 6 foundation deliverables are big swings; once 2-3 land, velocity catches up.
- **accountability_hud: MEDIUM ENTROPY.** Active bug fixing (2 P0 bugs) + 4 supervised feature requests on HUD-off-limits surfaces. Healthy cluster, just gated on jimbo's bandwidth.
- **mgp, playbooks, pm_system, daily_stop, monthly_reports: LOW ENTROPY.** Active, shipping, well-scoped.
- **live_classifier, mtf_vp_scanner, param_surface: LOW VOLUME but healthy.** Small surface area.
- **alerts-hotkeys-etc: NEW, INACTIVE.** 1 task created 5/12; not yet meaningfully attended to. Watch but don't act.

## Active-attention overflow check

Per Pass 9 spec: "active-attention overflow (>4 properties with >3 P0/P1)".

| property | P0/P1 pending |
|---|---|
| scorecards | 16 |
| foundation | 11 |
| accountability_hud | 8 |
| pm_system | 6 |
| mgp | 4 |
| param_surface | 3 |
| risk_hud | 4 |
| strategy_library | 4 |
| playbooks | 2 |
| live_classifier | 1 |
| mtf_vp_scanner | 1 |
| daily_stop | 1 |
| monthly_reports | 0 |

**8 properties have >3 P0/P1.** That's double the warning threshold. James is splitting attention across most active surfaces. Confirms the prime directive's bias toward leverage-per-attention-unit and supports compounding-infra prioritization.

## Composite entropy

| Tier | Properties |
|---|---|
| HIGH | scorecards, risk_hud |
| MEDIUM | foundation, strategy_library, accountability_hud |
| LOW | mgp, playbooks, pm_system, daily_stop, monthly_reports, live_classifier, mtf_vp_scanner, param_surface |
| INACTIVE | alerts-hotkeys-etc |

## Cluster signals (cross-task)

Tasks within 2 SD of each cluster centroid:

1. **Trader-picker cluster** (shared-user-dropdown, multi-trader-scorecards, phase-4-trader-selector, monthly-per-trader-view): 4 tasks all needing the same identity + dropdown component. Consolidation candidate (see Pass 9). All gate on `p0-identity-fn` + `add-user-role-approval`.

2. **Strategy codification cluster** (strategy-vpi-rules, strategy-vpi-backtest, strategy-backside-rules, strategy-backside-backtest, mgp-strategy-templates, strategy-align-scorecard-pie, strategy-followed-plan): 7 tasks all dependent on registry codification. Two of the seven (VPI rules + BackSide rules) are leaf nodes — the rest gate on them.

3. **acct-v2 promotion cluster** (acct-v2-services-task-scheduler, acct-v2-fills-poller-bridge, acct-v2-kore-tick-bridge-forward-test, acct-v2-positions-history-seed, acct-v2-retire-v1-marks-fast-daemon, acct-v2-recent-migrations-doc-delta, acct-v2-trade-action-requests-feature, acct-v2-promote-dev-to-prod): 8 tasks, 7 owned by jimbo/user. Coherent but blocked on jimbo's queue.

4. **Webhook enrichment cluster** (p0-indicator-alerts-ddl, audit-webhook-coverage, wire-webhooks, remove-na-labels, smb-rubric-grading, overtrading-detect, per-trade-charts, strategy-followed-plan): 8 tasks gated on user-owned p0-indicator-alerts-ddl. Strongest unlock-tree on the board (5 downstream from wire-webhooks alone) but requires James to run psql.

5. **Foundation P0 cluster** (p0-freshness-dashboard, p0-orchestrator, p0-daily-summary-view, p0-identity-fn, p0-param-surface-registry, p0-indicator-alerts-ddl): the system contract phase-0 list. 4 claude-owned + EXECUTION_READY, 1 blocked on user migrations, 1 user-owned. **Highest aggregate leverage on the board.**

## Summary

- **2 HIGH-entropy properties** (scorecards, risk_hud) — both need consolidation/decisions, not more work.
- **8 properties with >3 P0/P1** — confirms attention overflow; supports compounding prioritization.
- **5 cross-task clusters** identified — Pass 9 acts on the trader-picker + strategy-codification + foundation P0 clusters.
- **Composite entropy verdict: MEDIUM** at the system level. The system is shipping (101 completed in 30d, healthy nightly cadence) but the open queue has grown 17% (110 today vs the 94 it would be without the 16 priority-overflow tasks beyond the ideal 4-property attention window).
