What people ask before they join.

No. It's a layer before — you simulate dozens of scenarios and take only the survivors into fieldwork. Real research gets sharper focus, not eliminated.

No. They are cognitive architectures: the demographic marginals are calibrated on official statistics, and the behavioral axes are authorial, with adherence to that profile measured across five lots and three model families. We never model an identifiable individual. The unit is the cohort, not the person.

Every output arrives as an uncertainty band, not a single number, with a provenance card listing calibration sources. A scientific instrument, not a crystal ball.

Because it's the hardest test. A model that holds up in markets as heterogeneous as Brazil's is more honest than one calibrated on a homogeneous audience. The cultural density here is our calibration, not our boundary.

It answers from day one. Attach your own material and report real outcomes — the confidence seal evolves through backtests, and it's never green without one.

Predictive policing, partisan electoral targeting, pricing that violates LGPD, models of workers without consent. The refusal ledger is public — "no" is part of the product.

Through the app at app.syntheticperson.ai — create an account and sign in. The product is in beta; larger teams can talk to us through the form.

Measured, not claimed. Against a real Brazilian survey of 1,557 people, the current ruler lands at a mean absolute error of 16.73 percentage points across 120 questions — above the 15-point threshold, so that population's seal is Directional, not Calibrated. Use synthesis to discover and rank hypotheses; use real measurement to confirm the one you will bet money on. The full ledger, per segment, is public.

A panel measures one wave; here the next scenario costs minutes, not a new field. Every output arrives as an uncertainty band with its provenance, and per-segment validity is declared before you read the number. The omnibus measures; this discovers — and sends the survivors to the field.

There is no public price today. The product is in beta at app.syntheticperson.ai: create an account, and every cost is shown and approved by you before a study runs. Larger teams and partnerships go through the contact form.

No identifiable individual is ever modeled — the unit is the cohort. What you attach stays in your account and moves the prediction, never the confidence grade. In dialogue, requests for personal data are blocked at the input and any formatted identifier is masked at the output. Contact details you give us are used only to reply, under Brazil's LGPD.

No. Partisan electoral targeting is on the public refusal ledger, alongside predictive policing, pricing that violates LGPD and modeling workers without consent. "No" is part of the product.

The architecture runs wherever official public statistics exist. What does not travel is the seal: every country earns its own by measurement, never by transfer. Today the calibrated library is Brazil — which is also the hardest test, and the reason it starts here.