Daily Writeup — Monday, 2026-07-27

Monday · Generated by nightly-project-manager · ← All Todos · ← Jul 26
7
Shipped
3
Moved Forward
4
Blocked

TODAY — Top 3

  1. Fix the nightly-PM scheduler — it never fired 7 nights running (7/20–7/26). The startup heartbeat log has a LAUNCHED line for 7/19 and then nothing until tonight: zero launches for seven consecutive nights. This is the upstream scheduling daemon, not a mid-run crash — there's no partial output either. Tonight's fire had to backfill a full week by hand. Until the daemon is fixed, the todo list, the progress calendar and the daily writeups all go dark again the moment this machine sleeps through 23:00.
  1. Confirm Execution Scorecard v3 is the canonical scorecard and retire the pre_v3 sprawl. v3 landed all week: first injection 7/24, effective-stop enrichment + v3_eod.sh 7/25, history backfill 7/26, and today v3_live.sh went headless and run_all_eod.bat was rewired to call it. What is not settled is which page James actually opens — every date now has both trade_scorecard_<d>.html and scorecard_v3_<d>.html plus a .pre_v3_inject.bak. Pick one front door, point manifest.json and nav at it, and clear the ~40 .bak files before they become the next stale-twin problem.
  1. LAND the freshness-monitor decision — make atlas-flow-FULL canonical, retire app-status.json. Sixth night on this list. app-status.json is ~19 days dark (frozen 2026-07-08); atlas-flow-full refreshed cleanly again tonight at 20:40 and is the proven live monitor. Retire app-status.json and the content-stale atlas-flow-status twin, repoint the memory note, wire [WATCHDOG] at the live one. Note the irony: the nightly-PM outage above is exactly the class of silent failure this monitor exists to catch.

What shipped

What moved forward

What's blocked

Files touched

Generated by nightly-project-manager — 7 shipped, 3 moved forward, 4 blocked