α
AlphaSteve
← Brain index

Thesis Pre-Flight Checklist — what every thesis runs before publishing

The five gates in 05-decision-framework specify what a thesis must demonstrate. The deliverable-suite specifies what artifacts a thesis must produce. This file is the operational checklist that confirms both before the verdict is published — the last item between a draft and a verdict that routes a name into the portfolio, watchlist, or filed.

The point is not bureaucracy. The point is that operating discipline degrades fastest in the first three months of live operation, when no calibration tracker has yet appended and no kit-debrief has yet pointed at a sloppy artifact. A checklist enforces the standard while the operator's instinct is still forming. After 10–15 cycles, the checklist becomes automatic and largely redundant; in the meantime, it is the cheapest insurance against the failure modes the rest of the kit was designed to prevent.

A thesis that does not pass every box on this checklist is not yet a thesis. The draft remains in the workbench; the verdict is not published; the name does not route to portfolio or watchlist.

Pre-flight — the five-gate confirmation

The decision-framework gates, restated as binary checks the operator confirms in writing in the thesis's frontmatter or in a pre-flight section at the top.

  • Gate 1 — Circle of competence. Can I describe how this business makes money in one paragraph, without jargon? Can I name the three or four economic variables that drive its long-term value? Have I read at least one sector file from 07-Sectors/ and at least one moat / bottleneck file relevant to this business?
  • Gate 2 — Knowability. Does the thesis rest on questions that are answerable with effort (assets, normalized earnings, moat durability), or does it rest primarily on an unknowable (binary regulatory outcome, single-quarter revenue print, takeover speculation)? If the latter, is it sized as a bet rather than an investment, with documented EV math?
  • Gate 3 — Valuation triangulation. Have I produced at least three valuation methods and stated a range with a midpoint? Have I cited EPV (Greenwald) explicitly per earnings-power-value-greenwald? Is the asset / liquidation floor identified per liquidation-and-asset-value?
  • Gate 4 — Why is it on sale. Have I stated the reason consensus is wrong using one of the six honest answers (time arbitrage / forced selling / behavioral / information edge / variant interpretation / complexity discount)? Or have I documented why I have concluded the market is right and I am missing something?
  • Gate 5 — How do I get killed. Have I written three to five kill criteria as specific observables (not "if it goes down a lot")? Have I stated the top three downside scenarios with rough probabilities and the dollar-loss in each? Have I sized the position consistent with the range width and the binary risk per position-sizing-kelly?

If any box is unchecked, the gates have not been passed. Returning to drafting is the correct outcome.

Tier 1 deliverable confirmation

Every thesis ships with the three mandatory artifacts per deliverable-suite. The checklist confirms each exists and is complete, not just present as a placeholder file.

  • <TICKER>.md — the thesis note. All eleven required sections from the investment-thesis-template are populated: bottom line, business, structural assessment, variant perception, valuation triangulation, quality and management, macro/cycle, risk/pre-mortem, position sizing, what I don't know, next review.
  • <TICKER>-dashboard.html — the dashboard visualizer. Renders the recommendation pill, the four key numbers, the valuation triangulation chart, the reverse-DCF sensitivity grid, the scenario probabilities, the kill criteria, and the signal callouts. Opens without error in a browser. Used as the 60-second re-read anchor at every calibration checkpoint.
  • <TICKER>-calibration.md — the calibration tracker. Copied from _calibration-template. Initial prediction snapshot filled: price at verdict, central value, scenario probabilities. Specific testable predictions with timelines listed. Kill criteria copied with empty checkbox state. Re-read schedule (quarterly + 12-month + 24-month) populated with dates.

Source discipline confirmation

The sources-policy is the constitution for evidence. Pre-flight confirms it has been applied to the draft.

  • Every load-bearing claim carries a tier tag in the form <span class="tier-t1" title="source">T1</span>, <span class="tier-t2" title="source">T2</span>, or <span class="tier-cal" title="...">AS-cal</span>. Skim the draft and count: are there any unsourced claims about numbers, market structure, management behavior, or industry dynamics?
  • No Tier 4 sources. No Wikipedia. No generic LLM output. No anonymous Substacks or unsigned Seeking Alpha pieces. If anything trace-back fails this test, it is removed or downgraded to opinion explicitly labeled as such.
  • T1/T2 ratio is reasonable for the conviction tier. Core 1 / Core 2 theses should be predominantly T1-backed on the load-bearing claims. A buy verdict with the variant perception resting only on T2/T3 sources is a thesis that hasn't done the primary-source work the soul requires.

Variant perception confirmation

The variant perception is the section the rest of the thesis exists to support. The checklist makes sure it is actually present and falsifiable.

  • Consensus is stated as a specific claim, not a vague sentiment. "Consensus mean PT is $X, implying revenue CAGR of Y% and terminal multiple of Z×" — not "the market is bullish."
  • Our view is stated as a specific claim that differs from consensus on a named variable. "We see revenue CAGR of Y′% because of [observable]; we see terminal multiple of Z′× because of [structural reason]."
  • The basis for the divergence is identified. Are we relying on different inputs, different weights, or different time horizons? Each implies a different test for whether we are right.
  • The thesis identifies what evidence would converge our view toward consensus, not just what would confirm it. This is the inversion discipline from pre-mortem-and-inversion.

Verdict routing confirmation

Per deliverable-suite, the verdict determines the downstream container.

  • Verdict is explicit and uses the four standard values: buy / pass-with-trigger / pass / avoid.
  • Verdict reasoning is one paragraph explaining why this verdict and not the adjacent one. A buy that doesn't articulate why it isn't a pass-with-trigger is missing a discipline check.
  • If buy: position size tier (Core 1 / Core 2 / Mid / Probe) is named, with one-line justification anchored in conviction × range width. Hard limits per Portfolio confirmed (no >15% at cost, no industry >30%, no country >25%).
  • If pass-with-trigger: trigger price is specified, central value is specified, gap-to-trigger is computed. Re-engagement protocol from Watchlist understood and applied.
  • If pass / avoid: the verdict reasoning makes clear what would have to change for the name to be re-engaged, and the thesis is filed in the workbench rather than left in draft.

Operating-state confirmation

The thesis is not produced in isolation; it is produced in the context of the current portfolio and the current _house-view. The checklist surfaces this context.

  • House view reconciliation: the thesis has read the relevant sections of _house-view (sector, macro, regime) and either confirms the standing position, extends it, or flags a conflict explicitly.
  • Drawdown-state check: if a drawdown-protocol band is currently active, the thesis confirms compliance with that band's restrictions. A buy verdict at Band 2 (−10%) requires the documented override per the protocol. A buy verdict at Band 5 or 6 is forbidden.
  • Opportunity-cost check: if this is a buy verdict, the thesis names which existing position (if any) it displaces or which cash position it consumes, and why this opportunity is more attractive than the alternatives.

Sign-off

When every box is checked, the thesis is ready to publish. The verdict is recorded in the frontmatter, the deliverables are committed to the per-thesis folder under 09-Theses/<TICKER>/, and the routing happens:

  • Buy → row added to Portfolio, transaction logged in Transactions, drawdown-state header updated if material
  • Pass-with-trigger → row added to Watchlist with trigger and central value
  • Pass / avoid → filed in workbench, no further routing

The pre-flight checklist itself is appended to the thesis as a frontmatter field or short section showing the checks were performed. The audit trail matters: at the six-month-test checkpoint and at every kit-debrief, the existence of completed pre-flights for the closed-thesis set is one of the things scored.

What pre-flight is not

  • Not a substitute for the work. A checklist of completed steps is not the same as a thesis that passes the five gates substantively. The boxes are confirmations of completed work, not replacements for it.
  • Not a guard against being wrong. A thesis that passes every check can still produce a losing position. The checklist guards against avoidable procedural failures, not against the inherent uncertainty of equity work.
  • Not static. As the calibration record accumulates patterns of missed checks correlating with bad outcomes, items get added or refined per the Tier 2 process in Rules.

Linked