Nice narrative. Where are the numbers?

How the analysis begins: audit the books before the story

The Michael Burry Agent's default posture isn't "why does the market love this?" It's: is there a line in the 10-K or 10-Q quietly contradicting the story? Revenue is growing — are receivables growing faster? Earnings look great — is the cash keeping up? Are stock comp, depreciation, and capitalized costs quietly pushing real expenses into the future?

Its job is not to make a long-or-short call for you. It tests the company against accounting quality, the capital cycle, and valuation. Management may tell a convincing story, but the filings show whether the economics support it.

What it checks first

The agent prioritizes the latest available filings, market data, reports, and indexed public commentary. When newer evidence conflicts with older material, it uses the newer evidence and states the relevant date.

• Fresh data first — never numbers from memory Current prices, moves, market caps, rates, and filing figures all get looked up. Historical levels in old articles are anchors for their publication date, not answers for today.
• Primary sources over secondhand summaries It searches the long-form Cassandra Unchained essays first, then Short Thoughts, Trading Post, and X posts. Framework documents supply method and voice — they don't substitute for fresh data.
• Web search is a data layer, not an opinion layer News, announcements, analyst reports, and company talking points are raw material only. It extracts the data points first, runs them through the forensic-accounting and cycle frameworks, and never dresses a judgment up as "according to recent reports."
• Gaps in the source material are stated clearly If the available material does not cover a company, the agent labels the result as a framework-based analysis rather than implying that a published conclusion exists.

Where the forensic accountant looks

The forensic review focuses on the accounting lines most likely to expose a gap between the story and the economics: accounts receivable (AR), days sales outstanding (DSO), stock-based compensation (SBC), depreciation, inventory, deferred revenue, capitalized costs, construction in progress (CIP), off-balance-sheet purchase obligations, backstops, and residual value guarantees (RVGs).

If revenue is growing but DSO is rising even faster, the agent does not stop at "strong growth." It asks whether the pattern reflects channel stuffing, weaker pricing power, or revenue pulled forward from future periods. Management language is checked against the underlying numbers.

Valuation isn't a number — it's a set of implied conditions

On valuation, it translates the market price into implied assumptions: fifteen-year intrinsic value (IV15), the AI capital-cycle tracker (AICT), where price sits relative to IV15, the direction of earnings, the speed of cash conversion, and what the current narrative requires the company to keep pulling off, quarter after quarter.

When it cites valuation figures from past articles, it attaches the article date and reminds you those numbers belong to that moment's data — rerun them with current filings and prices. It won't hand you a lone "price target" dressed up as precision. It cares how much growth, how much margin, and how much return on capex that price is demanding — and whether those conditions have started to slip.

The capital cycle: where is risk building?

For AI data centers, semiconductors, and power infrastructure, it first identifies the stage of the capital cycle: early build-out, peak investment, or emerging oversupply. Hyperscaler capex, order books, delivery schedules, capacity expansion, financing conditions, and customer concentration show where risk is accumulating across chips, racks, networking, power, land, and balance sheets.

For companies such as Nvidia or Palantir, and for broader semiconductor exposure, the framework asks who funds the investment, who recognizes revenue first, who absorbs depreciation, and who ultimately converts demand into cash. These details matter more than an open-ended demand narrative.

What the output includes

A complete answer stays focused on four blocks rather than sprawling into an industry survey. The goal is a research conclusion you can verify, challenge, and keep tracking.

1. 01

   The forensic accounting verdict

   First, what's off in the books — or where nothing looks obviously wrong yet. The checklist: days sales outstanding, cash flow, stock-based compensation, depreciation, inventory, deferred items, and capitalized costs.

2. 02

   Capital cycle vs. the narrative

   Next, is it in build-out, at the peak, or rolling over — and has the market's story already outrun what the balance sheet, cash conversion, and industry supply-demand can support?

3. 03

   The transmission chain

   Then trace the risk or opportunity through customers, suppliers, financing, inventory, capacity, and pricing, to locate which link benefits first, strains first, or cracks first.

4. 04

   Asymmetric setup

   Finally, it identifies what would activate the thesis, what would invalidate it, and which variables matter next. When the risk-reward is not compelling, the answer says so plainly.

Try asking

Specific questions produce better work. Give it a ticker, an accounting concern, or a capital-cycle hypothesis so it can examine the filings, locate the cycle, and define the evidence that would change the view.

Who it's for

It is designed for investors who already have a company, supply-chain question, or long-short thesis and want to test the narrative against filings, cash flow, and capital-cycle evidence.

• US equity researchers who need to take apart 10-Ks, 10-Qs, cash flow, and receivables quality
• Investors testing whether the bull story is already more than priced in
• Supply-chain watchers tracking AI compute, data centers, chips, and power bottlenecks
• Anyone who wants a short thesis broken into triggers, invalidation conditions, and variables to watch

Boundaries: no compelling setup, no trade

The Michael Burry Agent does not promise returns, predict exact prices or timing, or encourage investors to use leverage or short a stock impulsively. Often, the disciplined response is to reduce exposure, stay on the sidelines, or wait for a more asymmetric setup.

It does not recommend options mechanically. Any hedge or options structure requires its own analysis of cost, duration, payoff, and risk. When there is no clear trigger, the agent identifies the evidence worth waiting for.

A short thesis is not a recommendation to short. The framework may identify downside risk, but the decision, timing, sizing, and loss limits remain the investor's responsibility.

Michael Burry

Bring a ticker. The agent will audit the financials, place the business in its capital cycle, and identify the assumptions the market may be missing.

Built from public materials to demonstrate a Burry-inspired research framework. It does not represent Michael Burry or provide investment advice.

Run $PLTR through your forensic accounting framework How far has the AI data-center capital cycle progressed? What's your logic for shorting SOXX?

Ask Michael Burry

Send