causalatlasResearch workspace
Interactive prototype · Synthetic dataFA
Research workspace / Scenario lab

Trace the impact of every scenario.

Explore how the same shock behaves across economic regimes. Trace the assumptions, factors and portfolio impact.

Example scenario loadedUSD · Synthetic research portfolio
Preset scenario parser · No LLM connected
SCENARIO TRANSMISSION / US MARKETS

Follow the shock.

Explore how an assumed macro shock reaches your portfolio.

CONDITIONING CONTEXTProbability blendSynthetic pre-shock beliefsCompare regimes
Ready to propagate15 nodes · 29 connections
Current portfolio——
0%
Assumed mechanismFactor mappingHolding exposureAggregation
100%
Drag nodes to arrange · Click a node or connection to inspect · Scroll horizontally on small screensAnimation explains model order, not market timing. All transmission parameters are synthetic.
Select a node or link to inspect its basis

Current portfolio analysis

All panels use the selected regime response and your current holdings.

Scenario return
—
—
Annualized volatility
—
Baseline risk · —
Tracking error
—
US balanced example benchmark · Annualized

Scenario transmission

Probability blend

The macro shock is an input; market responses are preset assumptions. Pricing uses one 10Y-based rates factor. The 2Y yield move is context only.

04

Multi-factor analysis

Synthetic factor model

05

Current holdings

Asset / Example holdingWeightScenario returnReturn contributionSimulated P&L
06

Portfolio what-if

Illustrative reallocation

Start by moving 10 percentage points of portfolio weight from long-term to short-term Treasuries. The actual transfer is limited to the long-term Treasury weight available.

10 pp
0 pp25 pp

About this research prototype

Explore assumed transmission from US financial events to a portfolio through economic mechanisms and risk factors.

01Choose a macro scenario, then run or replay the propagation. Pause or select a stage to inspect it.
02Adjust asset weights and compare a Treasury reallocation.
03Drag graph nodes, collapse holdings, or click a connection to inspect its assumptions and calculation. The animation represents explanation order, not real market timing.

Model limitations

Risk contributions use factor covariance and asset-specific risk. Scenario P&L uses factor shocks and portfolio exposures. Changing the shock size does not change baseline volatility.

The graph depicts assumed financial mechanisms and model mappings. No causal effect has been identified. Treasury yield and credit-spread responses are preset assumptions, not automatic consequences of a policy-rate change.

This financial network is separate from the execution topology an agent might use in LangGraph. LangGraph is not connected. TS-Agent-inspired structured evidence checks are mocked or computed locally; they are not an LLM execution trace or a causal proof.

The model is inspired by multi-factor risk analysis. There is no connection to BlackRock Aladdin, MSCI, live market data, or an empirical causal identification engine. FinNLI could support future financial-text inference; it would not establish causal effects.

Economic regime extension

Macro evidence packets and regime belief weights are fictional, manually specified examples. No economic-regime classifier is connected. The probability blend averages four preset response templates using those weights; selecting a regime conditions the same shock on one template. Exposures and baseline covariance remain fixed.

The input audit checks metadata and weight validity. Prepared ablation packets do not run LLMs, test historical memory, or establish out-of-sample performance. These evaluations remain not run.

J.P. Morgan’s Agentic AI for Dynamic Factor Investing (17 September 2026) informs the regime and ablation workflow based on a user-provided first-page screenshot. The complete report has not been verified. Regime-conditioned shock transmission is our prototype extension, not a reproduction of the report.

Methodology references

BlackRock: Making of a market-driven scenario ↗TS-Agent methodology paper ↗CFA research on causal investing ↗FinNLI: Financial natural-language inference ↗
These sources are methodological references. They do not validate this prototype’s parameters, estimated results, or specific transmission assumptions.

Evidence & assumptions

Export current analysis

Includes scenario settings, portfolio weights, factor results and supporting records. All content is marked as synthetic.

If the embedded browser does not save the download, copy the JSON instead.