Frequently Asked Questions

StressGen is a market intelligence and AI financial stress testing workspace for analysts and financial risk teams. Research market conditions, inspect searchable evidence, trace economic transmission channels and generate historical, hypothetical or black swan stress scenarios. Euclide, the Risk Assistant, helps investigate research questions and explore scenario ideas.

Ask Euclide to research recent news, companies and filings, explore economic implications, or answer follow-up questions about a scenario. Research stays in chat with source references. Euclide can propose a scenario for you to review and choose to generate; asking a research question does not automatically update a saved scenario or enable monitoring.

Rules & Alerts is an optional feature for enabled beta companies, separate from Euclide research. Rules are checked around the clock and each reading is as fresh as its source. Alerts follow notification preferences.

When the company-rules beta feature is enabled, company admins create rules, select an active seated admin as generation owner, test conditions, and explicitly enable monitoring. Each trigger can generate one company-private hypothetical scenario using the same shared allowance as manual generation. Members receive alerts according to their preferences. Alerts continue after allowance exhaustion; skipped scenarios are not generated retroactively.

Yes. StressGen automates financial stress testing end to end: describe a scenario in plain language and the AI pipeline produces a complete stress scenario — shock tables across 145+ risk factors, copula-validated secondary effects, narrative documentation, and portfolio P&L impact. Scenarios are AI-augmented rather than fully automated: every shock carries provenance and rationale so risk teams can review, challenge, and adjust before sign-off.

StressGen retrieves relevant context, compares economist proposals, constructs a scenario and applies statistical validation using copula models. Review cited evidence, assumptions and retrieval gaps in Market Intelligence; the Pipeline view exposes run status and diagnostics. Available sources and observation dates vary by scenario.

StressGen supports three scenario types: Historical (replay past crises like the 2008 GFC or COVID crash with historically-calibrated shocks), Hypothetical (forward-looking scenarios driven by live market intelligence, news sentiment, and prediction market signals), and Black Swan (extreme tail-risk events with low probability but severe market impact). Each type can be generated at baseline, adverse, or severely adverse severity levels.

StressGen covers 145+ risk factors across equities, interest rates, FX, commodities, credit spreads, volatility, inflation, macroeconomic indicators, and crypto. Primary factors (85) are proposed by AI economists, while secondary factors (47) are computed through copula-based correlation modeling to ensure statistical consistency.

StressGen supports three scenario types. Historical scenarios replay past crises (2008 GFC, COVID crash, eurozone debt crisis, 1994 bond rout) with historically-calibrated shocks. Hypothetical scenarios are forward-looking, driven by live market intelligence, news sentiment, and prediction market signals. Black Swan scenarios model extreme tail-risk events with low probability but severe market impact. Users can also upload their own documents into Knowledge Folders to ground any scenario in organisation-specific guidance.

StressGen integrates 12+ live data sources including FRED (Federal Reserve Economic Data), Yahoo Finance, ECB, Federal Reserve, multiple providers for news and web research, curated RSS feeds and GDELT for global news coverage, and Kalshi and Polymarket prediction markets for forward-looking indicators.

Coverage assesses retrieved context, signal relevance and source quality. Evidence alignment weighs signals supporting or contradicting the scenario; it is an index, not the probability that an event will occur. Confidence combines economist agreement and statistical validation when those inputs are available. Missing components are shown explicitly, and older scenarios retain their original scoring methodology.

Each AI-augmented stress scenario passes through a multi-stage pipeline: context gathering from 12+ live data sources, an economist debate that proposes primary shocks with cited rationale, copula-based propagation of secondary shocks, and a consistency check before output. The result is a financial stress scenario with a full audit trail — every number traceable to the data and reasoning that produced it.

Three AI economist agents with distinct perspectives propose primary shock values through a structured debate of up to three rounds. Each economist justifies their projections using real-time data. A consensus mechanism synthesizes their proposals into coherent, defensible scenario parameters.

StressGen uses Archimedean copulas (Clayton, Gumbel, Student-t) to compute secondary risk factor shocks that maintain realistic statistical dependencies. When primary shocks are set by economists, copula models propagate consistent secondary effects across credit spreads, volatility indices, and emerging market currencies.

StressGen pulls real-time contract prices from Kalshi and Polymarket to capture market-implied probabilities of economic events. These forward-looking signals complement traditional macro data and news, providing a crowd-sourced perspective on potential stress scenarios and their likelihood.

A grounded scenario connects its market observations to source evidence while keeping hypothetical assumptions and modeled consequences explicit. StressGen combines relevant macroeconomic indicators, market data, news, prediction markets and knowledge documents. Citations support observations; they do not prove a hypothetical scenario will occur.

StressGen uses AES-256 encryption at rest and TLS 1.3 in transit. Multi-tenant isolation ensures organizations cannot access each other's data. All personal information is automatically stripped via PII sanitization before reaching any LLM provider. StressGen never sells or shares customer data.

No. StressGen automatically sanitizes all personally identifiable information before any LLM call using a dedicated PII sanitizer. Names, account numbers, and other sensitive fields are stripped or replaced with safe placeholders, ensuring no client data reaches external AI providers.

All data is encrypted with AES-256 at rest and TLS 1.3 in transit. Database connections use SSL with certificate verification. API keys and secrets are stored in encrypted environment variables, never in source code or client-side storage.

StressGen is licensed annually per company. Pricing depends on your team, use case and generation capacity, so we quote it directly rather than publish a list price. No payment card is required to request private beta access.

Yes, premium data sources are available on demand. Contact sales to discuss your needs.

No payment card is required to request private beta access. Administrators provision company capacity and assign seats, and your team shares the company generation allowance. Checkout remains paused.

Starting a generation reserves one unit of company allowance. Failed, cancelled, and output-degraded generations release their reservation. Successfully completed generations consume one unit.

Request private beta access with no payment card required. An administrator provisions your company and team seats. Describe a stress scenario in plain language to generate a context-grounded scenario using your company’s shared monthly allowance.

Yes. Custom data feeds, on-prem deployment, dedicated support, and custom pricing libraries are available through Enterprise contracts. Contact sales for institutional terms.