Simfolk

Digital twins that stay grounded in real people

Simfolk builds digital twins of real humans—models trained on behavioral data and connector signals so each twin simulates how that person thinks, reacts, and decides. It helps teams evaluate launches, positioning, and tradeoffs before investing in production.

Instead of generic personas, you get an advisor cohort grounded in real user context. Run simulations, inspect disagreements, and use Past Runs to compare outcomes over time.

Trust signal

Real user-grounded behavioral signals

Response time

Waitlist and access requests handled within 12 hours

Workflow

Connectors hub for setup, Simfolk hub for simulation

Run SimulationsView Run HistoryOpen Connectors HubJoin Waitlist
CogVexa

Common use cases

  • Pre-launch product testing across realistic audience slices
  • Board-style strategic debate before real-world spend
  • Message and positioning stress-tests with memory-grounded twins
  • Faster iteration cycles than surveys and static personas

For full setup and connector onboarding, open Simfolk Connectors.

Why Simfolk is different from Mirofish

Simfolk

Real user-grounded simulation signal

Mirofish

Generic AI survey agents

Trained on behavioral data from real users, not invented personas.
Synthetic personas generated from prompts without real behavioral grounding.
One dedicated model per person so each twin keeps a unique behavioral fingerprint.
Single model prompted to "act like" many personas.
Returns structured cohort-level signal you can compare and act on.
Text-heavy persona outputs that are harder to aggregate into decisions.
Built for statistically meaningful panels of real users.
Usually a smaller synthetic set chosen by prompt design.
Designed for pre-launch validation with real potential users.
General-purpose persona simulation that may miss your actual user base.
Your models and signal stack are built around your own data graph.
Cloud product abstraction where internals are opaque.
Gets sharper over time as real behavioral data grows.
Personas stay static unless prompts are manually reworked.

The core difference

Mirofish asks AI to pretend to be people. Simfolk trains on real people and lets their behavioral fingerprints speak for themselves. For products with real early users, Simfolk's signal is grounded in truth - Mirofish's is not.