Bring us a question worth answering.
We work with labs, institutions, and independent researchers who want to run studies on synthetic populations, and publish what they find. You bring the hypothesis and the domain. We bring the substrate, the methods, and a commitment to write up the parts that didn’t work.
Recruiting the right thousand people is the study.
Recruiting 1,000 people for a multi-month study takes months. It costs a fortune before the first survey goes out. Half of them quit by wave three.
The hard part of research was never the analysis. It was finding exactly the right 1,000 people, and keeping them to the end.
So we build the population instead.
Each resident is generated across 105+ psychometric dimensions, lives inside a social graph, earns and spends, and leaves a reasoning trace behind every decision.
105+ psychometric dimensions
Not a demographic label with a voice on top. A profile with a shape: how curious, how disciplined, how warm, how much risk they will actually carry.
A position in a social graph
Residents are places in a network, not independent draws from a distribution. Who someone knows changes what they do next.
Money that runs out
Every resident earns and spends. A stated preference that survives a budget is worth more than one that never had to meet a price.
A reasoning trace per decision
Each choice arrives with the sentence behind it, so a finding can be interrogated rather than taken on trust.
zeldaCity was the first one we published.
Since then we have been running several research tracks in parallel. The engine is ready for other people’s questions.
- 929
- Residents
- 150
- Simulated days
- 52,677
- Relationships, none of them scripted
An instrument tuned to your field.
A synthetic society tuned to your field
Personas built for the questions your field actually asks. If your work studies how people decide, it probably fits.
- Healthcare
- Economics
- Entrepreneurship
- Public policy
- Consumer behaviour
- Education
- Urban planning
- Organisational research
Stakeholder simulations
Talk directly to the exact audience your research is about, at whatever scale the question needs. Ten respondents or ten thousand, same afternoon.
Synthetic polling
Polling run across that society, tuned to your target segments and ICPs, with the crosstabs you would have had to buy a panel to get.
Methods, not just outputs
You get the generation pipeline, the validation approach, and the failure modes we already know about. A result you can’t interrogate isn’t a result.
Replication support
Seeds, configurations, and population snapshots, so a finding can be re-run rather than believed. We would rather be checked.
You run it blind. We check it against ours. We publish both.
- 01
You run your study blind
You never see our internal benchmarks while the study is live, so your results cannot bend toward them.
- 02
We validate alongside you
We collaborate with you on the design and validate your results against our internal tests once your numbers are locked.
- 03
We co-author the white paper
Including a variance study: zeldaLabs synthetic personas against real humans, on your questions and your study.
We know the critique of silicon samples. We share most of it. A blind study with your ground truth next to our output is the only honest way to answer it.
- That the work is published
- Not necessarily with us, and not necessarily favourably, but published.
- That the limits travel with the findings
- Simulation gives you the trend, not the market share. We say that in our own papers and we would want a partner to say it too.
- That we can say what we don’t know
- We are a small team working on young methods. If a study finds the instrument isn’t good enough for the question, that is a result we want to write down rather than bury.
The first cohort is open.
Built for research labs, university research wings, and corporate innovation teams with a validation problem.
Research labs and university wings
Computational social science, behavioural economics, psychology, and public health groups who want a population to run an experiment on before running it on people.
Independent researchers
People pushing on synthetic cognition, agent-based modelling, and LLM evaluation who need infrastructure they don’t have budget to build.
Corporate innovation teams
Groups with a validation problem: a policy, a message, or an intervention that has to be tested across a population before it reaches one.
Bring us the question.
We respond to every inquiry within 48 hours.