Frame the decision.
You provide a question, an audience, optional context, and an alternative to compare. Foundate uses those inputs—not secret audience data—to frame an exploratory exercise.
METHOD, NOT MAGIC
More perspective.
Less false certainty.
Foundate is built to widen your thinking.
Not to tell you it can predict the world.
Assemble perspectives.
The agent creates four or six explicitly fictional personas with different priorities and concerns. These are creative lenses, not digital twins of real people or a statistically representative sample.
Make room for friction.
A language model simulates a moderated discussion. Personas challenge the proposal, react to each other, and consider an alternative. All voices come from one underlying model, so their apparent agreement is not independent evidence.
Find a next step.
The agent synthesizes tensions, assumptions, option trade-offs, and a concrete experiment. Ask follow-up questions to examine the result. Export a Markdown brief or revisit saved sessions on this browser.
PLEASE KEEP THIS IN VIEW
An input to your judgement.
Never a substitute for it.
- No live research. Foundate does not browse, survey people, retrieve current facts, or verify the context you provide.
- No predictive validity. Outputs are generated hypotheses. They are not opinion polls, probabilities, market forecasts, or evidence of demand.
- Bias is possible. Fictional personas can reproduce stereotypes, miss lived experience, and share the underlying model’s blind spots.
- Keep humans in the loop. Validate with actual interviews, experiments, and qualified experts. Do not use it to make high-stakes medical, legal, financial, employment, or public-policy decisions.
- Keep private information private. Do not enter personal data or confidential material you cannot send to an external AI provider.
GOOD QUESTIONS
Is this connected to Artificial Societies?
No. Foundate is an independent project by James He. It is inspired by social simulation, but does not use or reproduce the proprietary systems, datasets, research validation, or enterprise services of Artificial Societies.
What makes this agentic?
Foundate follows a three-stage workflow: plan a diverse fictional room, simulate a discussion, and synthesize a decision brief. Each stage uses the preceding result. It can then answer follow-up questions using the brief. It cannot take external actions, independently verify facts, or run a real-world experiment.
What model does it use?
The default backend uses Google Gemini 4 Argon through RouterHub. The site owner can select another RouterHub Chat Completions-compatible model in the server configuration. The sample run is prewritten and clearly labeled; it makes no AI calls.
How are sessions saved?
Completed briefs are saved locally in this browser when you enable “Save on this device.” You can delete them from the studio. Temporary server session data enables the three-step workflow. Accounts are optional and support email/password or verified Solana message signatures. Accounts do not sync briefs between devices; private SQLite storage holds account identity only.