Methodology
The short version is on the How it works page. This one covers what the scan looks for, how the score is built, and what a lead contains.
What it looks for
Developments first, stocks second. Policy changes, government contracts and procurement, supply shocks, regulatory filings, elections and polling shifts, industrial-policy moves, prediction-market repricings, unusual retail or search attention, launch schedules, production changes — anything happening in the world that is one, two or three causal steps away from a public-market consequence most people haven't traced yet.
The scan runs across broad web research plus a set of structured signal sources: prediction markets, SEC EDGAR full-text search, government contract and regulation feeds, Reddit ticker-mention deltas and search trends, and specialist feeds for space and automotive. Those are treated as signals, not truth — anything important is verified independently before it is published.
How a development becomes an opportunity
Each candidate gets a short causal exploration:
- What actually changed, and why could it matter economically?
- What happens one step later? And another step later?
- Who benefits disproportionately, who loses, and who supplies the beneficiaries?
- What becomes scarce, what becomes cheaper, who is forced to spend money?
- Is the obvious beneficiary already crowded while something adjacent hasn't moved?
- Is there a public-market proxy for something investors can't directly own?
- Has a risk been removed, or a probability changed, without a corresponding repricing?
- Is there a dated catalyst that could make the market suddenly care?
For big trends the question is always: if the obvious winner succeeds, what must also happen — and which link in that chain could become valuable without already being the obvious trade?
Selectivity
There is no target number of leads per day; some days there are none. Each candidate is evaluated on its own, in isolation — nothing gets a boost for fitting a theme with other items, and an idea that was published before only reappears if something material has changed.
Rejected on sight: generic observations ("AI is growing"), stories everyone already knows, themes that have already repriced enormously, speculative links with no causal mechanism, ordinary daily market news, obvious mega-cap beneficiaries with no extra insight, and situations with no realistic public-market expression.
Facts vs. hypotheses
Sourced facts are stated plainly. Inferences — about what markets are pricing, what investors believe, or how something might play out — are labelled as Hypothesis: in the text. Prediction-market odds are treated as crowd expectations, not as "smart money". If you paste an item into a research tool, that labelling is what stops it from researching the speculation as if it were fact.
The score
Every lead carries a score out of 100. It is a rubric total, not a probability — it answers one question: how worthwhile is it to spend a serious research session finding out whether this situation contains a genuinely underpriced or asymmetric opportunity?
| Component | Points | What it measures |
|---|---|---|
| Mispricing plausibility | 25 | Is there a credible reason the market may not fully appreciate the development or its consequences? |
| Economic materiality | 20 | If the thesis is right, does it materially move earnings, cash flow, supply/demand or capital flows? |
| Evidence & confidence | 15 | How well verified are the critical facts? Primary sources score highest; rumor scores near zero. |
| Causal strength | 15 | How convincing is the chain from real-world change to economic consequence to market implication? |
| Investable expression | 10 | Is there a reasonably clean public-market way to express the idea? |
| Catalyst / timing | 10 | Is there a plausible mechanism for the market to notice within an understandable timeframe? |
| Underattention / novelty | 5 | Is this genuinely less obvious than ordinary financial news? |
| 85+ | Drop everything |
| 78–84 | Strong lead |
| 70–77 | Worth a run |
| 62–69 | Thin, but live |
| < 60 | Not yet |
Scores above 90 are uncommon; above 95, rare. Scores are not spread to create a range. Each lead's Thesis check section states its strength and its weak link.
Anatomy of a lead
Same structure every time, so you can skim it or paste the whole thing:
- Headline and summary — what happened, and the short version of where the opportunity might be.
- Setup — the sourced facts, with absolute dates.
- Opportunity — why the obvious reading is incomplete, the mechanism, and the labelled hypothesis.
- How it could play out — the causal chain, step by step.
- Questions worth asking — the 2–5 questions that would actually decide the idea. Usually the most valuable section.
- Where to look — industries, countries, commodities and companies, each with a one-clause reason.
- Thesis check — what is strong about the chain, and the thing that would kill it.
- Sources — dated links, primary where possible.
Run it yourself
The deep research prompt is on the How it works page.