Daily writeup — 2026-05-15

Friday · Generated by nightly-project-manager · ← All Todos · ← May 14
0
Shipped
0
Moved Forward
0
Blocked

End-of-day fire 23:05 ET (on schedule). This page replaces the morning stub written at 08:04 ET; it covers the full 5/15 workday.

The day delivered a substantive build burst between 12:00 and 17:30 ET — 8 trackable ship items across MGP, playbooks, strategy registry, and the new doctrine layer — alongside another day of MGP scheduler wobble that needs root-causing.

TODAY — Top 3 (for Sat 5/16)

  1. Cut over MGP to the independent pipeline — Saturday dry run for Monday morning. Today's tape_read + gen_mgp_intraday + mgp_template iteration cycle is the third consecutive day of pre-cutover hardening; every supporting module is in place. Saturday is the right day to do a Vital-as-fallback dry run so Monday's 05:48 ET fire goes live without surprises. (Carry from 5/14 + 5/15 — same call, now even more overdue.)
  2. Validate strategies.json family migration end-to-end. Migration applied 16:17 ET with a .pre-family-migration.bak backstop. Before Monday: confirm /strategies/ page family filter still resolves, scorecard symbol-row classifier still maps the renamed families, and classify_v2.py output still matches expected family tags. One file move away from rollback if any of those regressed.
  3. Root-cause the MGP scheduler drift. Today's recap fired 14:14 ET instead of 16:37 — ~2h25m early. Combined with 5/14's mid-day+recap silent failures and 5/15's morning ~2h45m late, that's two trading days of scheduler misbehavior across three distinct fire slots. No longer a one-off. Inspect schtasks /query for the MGP triggers + cross-check against notify.log timestamps.

What shipped (build burst 12:00–17:30 ET)

+ xle-2026-05-14, with companion msft-charts.js / xle-charts.js and per-symbol chart_data.json. Extends the 5/14 playbook set from AMZN+NVDA to the full single-name + sector-ETF group.

snapshot preserved at strategies.json.pre-family-migration.bak (78,419 → 78,769 bytes). Aligns the registry with the canonical family taxonomy referenced by UNIFIED-FRAMEWORK-v1.

/system/UNIFIED-FRAMEWORK-v1.md. Composes the existing 13 Properties into a coherent strategy / scoring / optimization theory. Explicitly does NOT introduce new surfaces or redefine any property — pure architectural spec.

scanner/{engine, index.html, scanner.json} (16:58–17:05 ET). Shared playbook tooling layer separate from per-symbol pages.

gen_mgp.py, _tape_smoke.py, tape_read.py, gen_mgp_intraday.py. Drove today's 14:14 ET chop-day-tagged recap (the early one — see What's Blocked).

to fill the 5/14 silent-failure gap. Page recovered; root cause of the cron skip still under diagnose.

may-2026/index.html, vwap-bands/index.html, vwap-bands/param-surface/ (15 symbol surfaces+heatmaps, 7 TOD windows, tod_comparison, tod_sd_expansion), james-monthly-reports/index.html, vwap-research/index.html. Auto-pipeline output; ran clean.

What moved forward

yesterday's scheduler regression.

(trade_scorecard_2026-05-15.html) — scorecard pipeline still healthy through the family migration (good sanity signal but does not substitute for the Saturday end-to-end ripple check).

config-before-edit-2026-05-15T13-15-42Z.json backstop — user-initiated HUD config tweak (not a SKILL action).

What's blocked

slots affected. 5/14 mid-day + recap silent failure. 5/15 morning ~2h45m late (08:35 vs. 05:48). 5/15 recap ~2h25m early** (14:14 vs. 16:37, with a second regen at 16:38). Task Scheduler triggers need a direct audit — this looks like trigger drift, not a one-off freeze. P0 task mgp-cron-skipped-5-14 stays open and is now the dominant risk because it threatens the Monday cutover plan.

scientific-layer migrations awaiting psql tunnel.

ticket-meter bug — no movement today; both stayed P0.

Files touched