α
AlphaSteve
← Brain index

Cadence & Overhead v2 — cut the bookkeeping in half

Adopted 2026-06-10 as a hard doctrine change (see changelog-2026-06-10-hard-changes). Where this conflicts with the prior cadence in am-pm-template or thesis-bundle-standard, this file governs; those files carry pointers here.

The problem this fixes: for a zero-position $10k paper book, the kit produces two research notes a day, four-methodology shadow matrices on every name (including triage-stage names that may never get a full thesis), three-benchmark daily tracking, an audit log, a backlog, and daily optimization notes. Much of the daily output is elaborate documentation of why nothing happened. The mission file's own one-sentence test — does this materially improve our ability to find, evaluate, or act on mispricings? — fails for a meaningful share of it. Overhead is not free: it consumes the compute and attention that should go to sourcing figures and red-teaming theses.

Rule 1 — One end-of-day note by default

The default research cadence is one EOD note, not separate AM and PM notes. <span class="tier-cal" title="2026-06-10">AS-cal</span>

  • The single EOD note covers the US session, post-close prints, and sets the next-day posture — folding the old AM and PM into one.
  • A second (AM) note is written only when something material is pending that genuinely needs a pre-open read: a watchlist name inside its proximity band, a portfolio position with a catalyst, or a major scheduled event (CPI, FOMC, a held-name print). Absent a trigger, no AM note.
  • The three discipline rules from am-pm-template (primary-source pre-flight, ≤30% price-action cap, plain-English constraint) carry over unchanged — this is about frequency, not rigor.

Expected effect: roughly halves routine note volume; the notes that remain are event-driven and therefore worth reading.

Rule 2 — Four-way shadow matrix only at continue-stage

The full four-methodology shadow-valuation-matrix (Klarman / Greenwald / Buffett / Mauboussin) is reserved for names that reach the full-thesis stage (a continue first-read).

  • First-reads compute EPV-only, plus an asset floor. That is enough to triage. A name that dies at first-read never needed four methodologies.
  • The four-way matrix is built once the thesis-builder picks the name up — which is also when the banded-valuation-standard confidence tags and flip test are applied.

This removes the largest single piece of speculative bookkeeping (full matrices on names that wash out in triage).

Rule 3 — Consolidate benchmark tracking

  • One benchmark-tracking surface: the Performance ledger carries SPY as the primary benchmark daily, with RPV/RPG (the value/growth split) updated weekly, not daily. <span class="tier-cal" title="2026-06-10">AS-cal</span>
  • The daily portfolio task stamps NAV, SPY, and any open-position prices from the canonical closes file (data-integrity-gate Rule 5). The value/growth attribution — which only matters over weeks, not days — moves to the weekly summary.
  • While the book is 100% cash, the daily portfolio note collapses to a single line ("100% cash; NAV flat; no triggers fired; nearest gap: ") unless a trigger fires or a band is crossed. The current multi-paragraph daily register note is overhead with no decision content.

Rule 4 — The artifact test, applied quarterly

Every recurring artifact (each note type, each ledger, each optimization sub-file) is reviewed quarterly against the mission's one-sentence test. Anything that cannot be defended as improving find/evaluate/act is archived. The audit log, backlog, and calibration trackers pass the test (they are the learning loop). Daily optimization notes and per-day register prose largely do not, and collapse into the weekly/event-driven cadence above.

What is explicitly NOT cut

Linked