The claims Synset is making, and the ones it is not.

No.

Synset generates synthetic cohort-level futures constrained by clinical, temporal, and statistical validation checks. These are not patient-specific predictions and not true future outcomes.

Certification means a synthetic trajectory passed Synset's internal validation criteria before release. These checks may include physiologic plausibility, transition consistency, trajectory coherence, metadata safety, and evidence grounding for text sidecars.

It does not mean regulatory certification, clinical validation, or a guarantee that the synthetic trajectory is a real future patient path.

No. Real-world EHR data is not perfect either.

Synset's goal is not perfect data. It is controllable, privacy-preserving clinical simulation with explicit quality checks, configurable missingness, and measurable validation boundaries.

LLMs generate plausible text. Synset generates structured patient trajectories first, validates them, and only then produces optional text sidecars grounded to the structured evidence.

The structured trajectory remains the source of truth.

Synset has internally validated support-certified acute-care trajectory generation across 96 sequential hourly transitions.

Chronic and outpatient trajectory modeling is under active development. Current chronic prototypes are calibrated and release-gated against withheld empirical longitudinal-prior constraints, but they are not yet held-out patient-future accuracy claims.