Thesis Evidence Auditor — Test Your Pillars
Thesis Evidence Auditor — Test Your Pillars Against the Data
A research tool that turns an investment thesis into falsifiable pillars and checks each against current public evidence — reporting SUPPORTED / WEAKENING / CONTRADICTED / INSUFFICIENT DATA. It describes the state of the evidence; it does not tell you what to do. No hype, no fabricated numbers, no buy/sell calls.
Category: #Finance #Research #Analysis #DueDiligence
Model: Claude Sonnet 5
🔭 Overview
Investors fall in love with the story they bought and stop testing it. The Thesis Evidence Auditor is the disciplined second read: it breaks your original reasoning into checkable claims and reports, pillar by pillar, whether the latest public evidence still supports each one, has weakened, now contradicts it, or is simply missing.
It is a research tool, not a decision tool. It describes the state of the evidence and the gaps in it. It does not issue buy, sell, hold, or trim conclusions, gives no price targets or stops, and does no position sizing — that is your call, and your licensed adviser's. Every audit is delivered as a self-contained HTML report — open in any browser or print to PDF, no dependencies, works offline.
📊 What It Can Do
1. A self-contained HTML thesis-evidence report
Give it a ticker and the original reasoning behind a position — and it writes a single HTML file covering:
- Thesis Pillars — your reasoning broken into 2–4 falsifiable claims
- Pillar Evidence Audit — each claim tagged ✅ SUPPORTED / ⚠️ WEAKENING / ❌ CONTRADICTED / ◽ INSUFFICIENT DATA, with the source data and date behind it, plus an uncertainty note
- Price vs Narrative — apparent divergences flagged as neutral, open research questions (never an inference about informed trading)
- Open Questions & Data Gaps — the public information currently missing on a weakening or unclear pillar, and the report that would fill it (no numeric triggers, no position guidance)
- Data Gaps & Limits — what's stale, missing, or uncomputable
2. Audit a whole set of theses
Paste several positions with their reasoning. It gives a neutral comparison of which theses currently have stronger vs weaker supporting evidence — a description of evidence quality, not a ranking of what to sell or keep.
3. Honest data gaps
Backlog, forward-guidance internals, and anything not computable from public data are marked INSUFFICIENT DATA — never invented.
✨ What Makes It Unique
- Evidence status, not verdicts. Where other tools push a buy/sell call, this one stops at the evidence: SUPPORTED / WEAKENING / CONTRADICTED / INSUFFICIENT DATA. You keep the decision.
- Falsifiable, claim-by-claim. It names the exact data, source, and date behind each pillar's status — and states the alternative interpretation.
- Uncertainty stays visible. Every conclusion carries a confidence/limitation note; nothing is dressed up as certainty.
- Disciplined scope. It declines execution questions (what/how much to sell) and continues with the research audit — a tool you can trust to stay in its lane.
🚀 How to Use
| You ask | You get |
|---|---|
I bought NVDA for the AI-capex thesis — does the evidence still hold? |
A pillar-by-pillar evidence audit: SUPPORTED / WEAKENING / CONTRADICTED / INSUFFICIENT DATA, with sources and dates |
Has the revenue-growth pillar in my thesis broken? |
A focused evidence check on that claim, with the data behind it and what would clarify it |
Audit these 4 positions' theses |
A neutral comparison of which theses have stronger vs weaker current evidence |
Reconstruct and test the thesis for this stock |
Inferred pillars (marked for you to confirm) checked against the latest public evidence |
Best input: ticker + the original reasoning behind the position. No reasoning given? It reconstructs the pillars and marks them "inferred — please confirm." Any market.
⚠️ Important Disclaimers
- Not professional financial advice. This Agent reports the state of public evidence for research purposes only. It does not recommend buying, selling, holding, or sizing any position. Consult a licensed financial adviser before making any financial decision.
- Research states, not actions. Outputs are evidence statuses (SUPPORTED / WEAKENING / CONTRADICTED / INSUFFICIENT DATA), never buy/sell/hold/trim calls, price targets, stops, or position sizes.
- Public data may lag. Fundamentals and news are pulled from free public sources and can be delayed; every report is timestamped "as of {date}". All conclusions carry uncertainty; uncomputable figures are flagged INSUFFICIENT DATA.
💬 Examples
Example 1 — Does the evidence still hold?
You: I bought this chip name for its data-center growth. Does the evidence still support that?
Agent: Pillar A (data-center demand drives revenue YoY >30%): ✅ SUPPORTED — latest quarter revenue YoY +40% (source, date). Pillar B (margin expansion): ⚠️ WEAKENING — gross margin down ~2pts QoQ on mix (source, date); alternative reading: one-off mix shift. Pillar C (design-win pipeline): ◽ INSUFFICIENT DATA — not disclosed. Evidence state: 1 SUPPORTED, 1 WEAKENING, 1 INSUFFICIENT (a plain count, no implication for any position). Open question on Pillar B: the current-quarter gross-margin figure has not yet been reported; the next quarterly filing would fill that gap...
Example 2 — Is a specific pillar broken?
You: Has the growth pillar broken?
Agent: The growth claim is ❌ CONTRADICTED by the latest print — revenue YoY turned negative (source, date). Uncertainty: a single quarter; a rebound is possible. What would clarify: the next one or two revenue prints. (I don't provide buy/sell/hold decisions or price levels — this is an evidence audit.)
Example 3 — Declining an execution ask
You: Which of my two stocks should I sell first, and at what price?
Agent: I don't provide sell rankings, position sizing, or price levels. Here's the neutral evidence audit instead: Thesis 1 currently has 3 SUPPORTED pillars; Thesis 2 has 1 SUPPORTED, 1 CONTRADICTED (sources, dates)... — a description of evidence quality, not a recommendation.
