Enter the intelligence field to activate the licensed score and live collective view.
PUBLIC METHODOLOGY
This page explains what the public agents and models calculate, what they cannot claim, and why the private architecture is deliberately excluded.
Observed data comes from supplied evidence. Synthetic data is generated for the demo. Assumptions are editable beliefs. Simulations vary those assumptions. Inference interprets outputs; it is not new evidence.
Mathematics tests robustness; Finance models economics; Strategy frames advantage and experiments; Marketing distinguishes theoretical market from accessible demand. They return structured outputs and criticism, never chain of thought.
Defensibility tests imitation and switching-cost assumptions. Macroeconomics exposes financing, demand, currency and supply-chain sensitivity. Both are deterministic public models, not additional agents, forecasts or approvals.
Save, compare, request evidence, mark promising, uncertain or reject. No action invests, funds, deploys or executes a company.
An MCP specification or limited prototype may help validate technical feasibility. It does not replace product engineering, founders, operations, sales or governance.
The company workbench parses eight labeled financial fields, detects disclosed phrases in an earnings-call excerpt, blends a public DCF with an EV/EBITDA cross-check and flags a supplied negative beta. It does not fetch, verify or recommend a security.
DASHBOARD READING
Start with the comparative score, then test whether evidence, risk, uncertainty and specialist agreement support the same conclusion.
A weighted comparison under current assumptions. It is not a probability of startup success.
High scores with weak evidence require validation before they deserve confidence.
Read probability, impact, uncertainty and mitigation together. Lower is generally preferable.
Higher values mean reasonable assumption changes may materially alter the conclusion.
Inspect each agent when their scores diverge; disagreement is information, not an error.
Compare, request evidence, mark promising, uncertain or reject. No score approves an investment.
DEMONSTRATIVE MODELS · VERSION 3.0.0
How visible criteria and normalized weights contribute to one comparative score.
The score is relative to this run and is not a probability of startup success.Hundreds or thousands of possible outcomes under editable uncertainty and a reproducible seed.
Synthetic distributions amplify the quality - or weakness - of their assumptions.How ranking changes when agent importance, criteria or tolerances change.
Interactive reweighting is exploratory; the tornado model provides the explicit one-at-a-time calculation.How prior confidence changes after a new piece of evidence under visible likelihood assumptions.
The posterior is conditional on public assumptions and is not an observed success probability.Downside, base and upside results under different operating assumptions.
Three scenarios cannot represent every plausible future or correlated shock.Risk probability, impact, uncertainty and proposed mitigation.
Ordinal risk values require human ownership, evidence and domain review.Opportunities that remain competitive across two objectives without collapsing every trade-off.
The frontier changes when the selected objectives or assumptions change.A straight descriptive trend through the current synthetic opportunities and each residual.
Association does not imply causality and a 6-12 item sample is not predictive.The score range produced by changing one public dimension at a time around the current value.
One-at-a-time analysis does not capture interactions between changing assumptions.Contribution per retained customer, customers required, adoption required and payback under current assumptions.
It is an illustrative unit-economics threshold, not a valuation or financial forecast.How often each opportunity remains first or in the top three across seeded weight perturbations.
SMAA-style perturbations test public weights only; they are not historical calibration.An ordinal value-of-information-style index for the next evidence request from each public agent.
The index is not monetary EVSI and does not authorize research, spending or investment.A transparent view of imitation exposure, switching-cost assumptions and possible structural advantages.
It is a deterministic context lens, not a fifth agent or proof that a durable moat exists.Sensitivity to financing, demand, currency, supply-chain and time-horizon conditions.
It does not forecast GDP, inflation, interest rates or foreign exchange and is not financial advice.A five-year DCF blended with an EV/EBITDA cross-check from user-supplied figures.
It does not verify filings or prices and is not a buy, sell or hold recommendation.Which records in a disclosed synthetic universe have a beta below zero.
Beta is backward-looking and unstable; a negative estimate is not proof of hedging or fair value.This laboratory produces model-based estimates from assumptions and synthetic or user-provided data. It does not constitute investment advice and does not predict startup success.
The private registry aggregates 1,970 profiles across 51 domains, selected by mission. This public Lab demonstrates only four representative specialists; individual private profiles, proprietary selection and orchestration logic, and private quantitative models are not included.