Crene is investment thesis review infrastructure. The platform decomposes macro, labor, AI transition, and investment thesis questions into falsifiable components using four frontier AI models (Claude, GPT, Gemini, Grok). The system produces three product types: thesis maps (binary decomposition into assumptions), factors (continuous decomposition into drivers), and scenarios (hybrid decomposition into coherent pathways). Cross model disagreement is the primary signal. Calibrated against 1,161 resolved events with a Brier score of 0.114.

How does AI prediction calibration work at Crene?

Calibration applies at the consensus layer, not at the underlying model layer. Crene queries four frontier models in isolation, computes a cross model median consensus, and tracks resolution against a tiered source allowlist. Every resolved event becomes a permanent calibration point. Per tier accuracy, per model Brier scores, and per domain accuracy are computed live and published at https://crene.com/methodology.

What is a Brier score?

A Brier score is the proper scoring rule for probability forecasts on binary outcomes. It is the mean squared error between the forecast probability and the realized outcome (0 or 1). Lower scores indicate better calibration. The no skill baseline is 0.25 (always predicting 0.5); a perfect forecaster scores 0.0. Crene publishes the consensus Brier score and per model Brier scores across the resolved event corpus.

Crene serves investment teams that need a repeatable thesis review process: hedge funds, asset managers, macro PMs, CIO offices, family offices, and thematic investment teams. A team brings one thesis it is actively debating. Crene maps what the thesis depends on, where models disagree, what changed each week, and what would force a rethink. Data and API access remain available as the proof layer behind the workflow.

What is the difference between an event, a thesis map, and a factor at Crene?

Crene has four product types. Events are base binary outcomes with scalar probabilities. Thesis maps decompose binary anchor questions into child situations. Factor maps decompose continuous variables into drivers. Scenarios map long horizon investment questions into coherent pathways across live systems including Empire by Default, AI Labor Transition, European Rearmament, and India as the Third Growth Pole.

Proof layer · JUN 2026

The proof layer behind Crene thesis reviews.

Events, thesis maps, factors, and scenarios. Binary events carry four-model probability histories, named resolution sources, and Brier scores once resolved. Thesis maps, factors, and scenarios extend the same scoring discipline into structured decomposition: 3 live thesis maps, 3 anchor factors with 347 active drivers, and 400 pathways across 4 active scenarios decomposed into 266 scored components.

Read the methodology →

…active binary forecasts

Across all categories. Repolled daily.

…resolved questions

Closed against named sources. Binary outcomes are Brier scored.

…broad corpus Brier

Lower is better. Broad resolved corpus, rounded. See methodology for leakage-controlled and macro-only subsets.

…directional accuracy

Consensus direction vs outcome.

Assumption structure

Raw count tells you coverage. Structure tells you whether that coverage is actually diversified.

Crene separates the number of questions from the independence of the assumptions underneath them. A scenario, factor, or cluster can track many components while still depending on a small number of shared drivers. Question density across domains is the first read on that structure. As the observation record deepens, Crene estimates how many effectively independent assumptions sit underneath each object.

Question density is a structural read, not a correlation estimate. It shows how tightly each object's questions cluster by domain. Crene does not publish exact effective-independence or covariance figures until the observation record is deep enough to estimate them reliably. The same rule applies here as everywhere else on the page: no number is promoted before the record can support it.

Corpus reconciliation

The data page uses the broad corpus view. Total resolved questions refers to the full Crene-native resolved set currently in the database. The headline broad corpus Brier is rounded and differs from the leakage-controlled benchmark and macro-only subset discussed on the methodology page. Model leaderboard sample sizes are per-model rows, so they may differ from the consensus benchmark.

What is in the corpus

Multi model forecasts

Independent probability estimates from four production models (Claude, GPT, Gemini, Grok) for binary forecast objects. No model sees another model output.

Events

Binary outcome questions across macro releases (CPI, NFP, PMI), central bank decisions, commodities, crypto, policy, and market events. Each resolved against a named source.

Thesis Maps

3 live thesis maps decomposing regime-level binary questions into 309 active assumptions: Warsh Fed rate path (100), AI economic realization (109), and AI content dominance (100). Each assumption is independently forecast and tracked through resolution.

Factors

3 anchor factors with 347 active drivers. Continuous state variables (S&P 500, US 10Y Treasury yield, US GDP attributable to AI) are forecast as probabilistic distributions with daily driver refreshes.

Scenarios

400 pathways across 4 active scenarios: AI Labor Transition, Empire by Default, European Rearmament, and India as the Third Growth Pole. 266 components spanning demographics, AI, labor, defense, currency, fiscal, trade, governance, and market structure.

Belief trajectories

Consensus probability is snapshotted daily for binary forecast objects. The full evolution of the forecast is preserved through resolution.

Per event Brier scoring

Each model is scored on resolved binary events. Per model performance is measured continuously, not curated. Continuous factors and long-horizon scenario structures require different calibration metrics as outcomes accrue.

Continuously growing

New investment thesis maps, factors, and scenario components expand the corpus alongside macro, policy, market, and AI-related event flow. Resolutions compound the calibration signal.

How resolution works

Every outcome is tied to a verifiable source. Crene resolves binary events against a tiered source allowlist. Resolution is automated, but the source for each outcome is named and auditable.

Tier A

Primary

Government statistical agencies, central banks, regulators. SEC filings, BLS releases, Fed statements, BEA, BoJ, ECB.

Tier B

Authoritative secondary

Major news wires reporting primary releases. Reuters, Bloomberg, AP, agency wire confirmations.

Tier C

Corroborating

Used only to confirm Tier A or B. No event resolves on a single Tier C source.

Why this matters

Events, thesis maps, factors, and scenarios form the evidence layer behind Crene's thesis review workflow. Four model surfaces feed one resolved-event calibration record. Structural maps remain separate from scored evidence until outcomes accrue.

Access

Download Sample CSV View JSON

Proof layer access. Resolved event archive, per-event 4-model breakdown, calibration history, and thesis map decompositions behind the private thesis review workflow. Use the form below.

Marketplaces. Crene is listed on Neudata, Eagle Alpha, and Monda for institutional procurement.

API. Programmatic access to the proof layer behind live and resolved thesis objects. API documentation.

The proof layer behind Crene thesis reviews.

Crene Data: Calibration Corpus

The proof layer behind Crene thesis reviews.

Crene Data: Calibration Corpus

The proof layer behind Crene thesis reviews.