Institutional AI Simulator API
Simulate spins up a panel of distinct institutional archetypes, a macro allocator, a quant, a sector manager, a credit analyst, a retail reader, and a regulator, each parsing the draft through its own lens, then computes consensus, deviations, and the questions the filing will trigger.
What it returns
The response returns the personas panel (each with comprehension, concerns, and questions), a consensus_comprehension, a panel_agreement, a deviations list (the members who understand it materially worse, the blind spots), predicted_questions ranked by how many members ask them, a models_live flag, a frontier_models list, a frontier_disagreement value, and the trust envelope.
Send the draft text and optionally a symbol, which pulls a top macro driver from the Context Graph for the allocator's question.
When to use it
Use Simulate when you want the analyst-room view specifically: where a room of institutional readers would converge, diverge, and get confused, and the exact questions to pre-empt. It is the killer capability inside the AI Readiness Report, exposed on its own.
Deviations flag any persona at least 15 points below consensus, sorted by the size of the gap.
Notes and honesty
The personas are synthetic deterministic archetypes, not any real institution's AI, and the tool never labels one as a firm it isn't. When frontier models are configured they join the panel under the same contract, honestly labeled, and their spread of comprehension is reported as an empirical misread-risk signal. This measures the clarity of public text; it does not predict how any specific named model will respond. Not advice.
- personas: six archetypes with comprehension and concerns
- consensus_comprehension and panel_agreement
- deviations: the blind spots below consensus
- predicted_questions ranked by demand
- frontier_disagreement: empirical misread risk when models are live
- trust envelope over the panel
curl -s https://api.lyramindos.com/v1/ai-readiness/simulate -H 'Content-Type: application/json' -H 'x-api-key: YOUR_KEY' -d '{"text":"Segment performance was mixed; leverage remained within covenants.","symbol":"ACME"}'Questions
Are the personas real firms' AI?
No. They are synthetic deterministic archetypes such as a macro allocator, a quant, and a credit analyst. The tool never labels one as a firm it isn't.
What is frontier_disagreement?
When live frontier models are configured they parse the same text, and the spread between their comprehension scores is reported as an empirical misread-risk signal. It is null offline.
Is this a prediction about a model?
No. It measures the clarity and completeness of public text. It does not predict or guarantee how any specific named model will respond, and it is not advice.