Every briefing on this site ends in a dated, falsifiable call. This page explains how one gets made: where the questions come from, what reads them, and how the score is kept.
Where a briefing starts
We watch a fixed set of sources: regulatory dockets, filings, industry data, the trade press. Anything that smells like structure in motion becomes a candidate question. Most candidates die in triage. The ones that survive share one property: a market's layers are about to trade profit, and the move is callable in advance, on a horizon measured in quarters rather than decades. A person makes the final pick. Machines propose; they don't get to decide what we stake the record on.
The theory graph
Underneath the briefings sits a versioned knowledge graph built from the foundational strategy literature: Christensen on disruption, Wardley on evolution, Goldratt on constraints, Baldwin and Clark on modularity, Arthur on how technology composes, Teece, Reinertsen, West, Hidalgo. The graph stores each framework's concepts, the mechanisms connecting them, and the sources they came from. Eleven frameworks are live today; the lens catalog lists them with their track records. When a briefing says "margin relocates to the constraint," that claim links to the exact concept behind it. The glossary is the public face of those links.
Eleven lenses, read separately
Here is the part most research processes get wrong: consensus forms too early. Ask five analysts to debate and you get one blended view, usually the loudest one. So the lenses never see each other's work. Each framework reads the same question in isolation, cites only its own concepts, and files its own reading: what it expects, what evidence would change its mind, what it wants researched. Evidence requests get fulfilled and readings get revised, but a lens only ever revises on evidence, never on another lens's opinion.
Disagreement survives all the way to the page. When Wardley Mapping argues a constraint gets circumvented and Theory of Constraints argues it binds, you read both, with the weight each deserves. The dissent is not noise we smooth out. It is the most useful thing we publish.
Weighted futures, dated resolvers
Synthesis turns the readings into a small set of scenarios that split 100% of our credence. Each scenario carries the mechanism that would produce it, and each briefing carries resolvers: dated, public checkpoints that settle which branch is happening. No "we'll see." A date and a condition, registered before the fact.
The diagrams are drawn from the same data, not by hand: the history and forward timelines from the event record, the Wardley map from the component model. If the data isn't there, the diagram doesn't exist.
How the score is kept
When a resolver's date arrives, the call resolves, hit or miss, and the scorecard updates whether the result flatters us or not. We score with Brier: the squared gap between stated confidence and what happened. Zero is clairvoyance; 0.25 is a coin flip. It punishes overconfidence.
Two ledgers feed the page. The pilot's 48 pre-registered calls, made by hand before the engine existed, keep their original dates and confidences (Brier 0.156 across the resolved ones so far). Engine-era calls join them as briefings publish. Every lens also keeps its own ledger; the leaderboard shows which way of reading the world earns trust.
Backtests are the reverse test: freeze a past date, re-derive the calls from only what was knowable then, score against what actually happened. They are labeled as backtests and kept apart from the live record, so hindsight never inflates the number.
The four questions
Eleven frameworks is a bench, not a sermon. In practice they answer four questions, and we still group them that way when we talk:
- The Map. Where does each layer of the stack sit in its life cycle? Wardley Mapping; Combinatorial Evolution of Technology.
- The Shift. Which way do profit pools move as interfaces harden or dissolve? Disruptive Innovation; Modularity; Dynamic Capabilities.
- The Priority. What single constraint decides where the next dollar goes? Theory of Constraints; Product Development Flow.
- Financial Ground Truth. What do the unit economics actually say? Throughput Accounting; Scaling Laws.
The complexity lenses, Complexity Economics and Economic Complexity, cut across all four: they say when increasing returns lock a winner in and when path dependence makes the past a trap.
People and machinery
An engine does the bookkeeping. It keeps the graph versioned, holds the lenses to their isolation, registers every call before publication, and refuses shortcuts: a claim with no concept behind it doesn't render. People decide what enters the pipeline, review what leaves it, and own every call the site publishes. If the record goes bad, that is on us, not the tooling.
Editions
Each briefing publishes at two depths. Signal is free: the call, the weighted futures, the reasoning that matters. Analyst is the full analytical edition: every diagram, the complete lens readings, the appendix with the working. While the track record builds, everything is open to read.