ZeldaLabs
TownSquare, our first product is live! Check it out →

We build
what's next.

From frontier research to shipped products. Tomorrow's AI, shipped today.

Applied AI Research

ZeldaLabs is an applied AI research company. We work at the frontier of synthetic human intelligence, building systems that understand how people think, decide, and behave.

The Problem

The gap is growing.

Frontier AI research is advancing faster than ever. But the distance between a breakthrough paper and a product people can use has never been wider.

We exist to close that gap. Turning the hardest unsolved problems in AI into tools that work at production scale. No eight-week timelines. Just answers.

01

Research stays locked in labs

Breakthrough papers pile up. But the path from publication to product is years long, if it happens at all.

02

Products stay frustratingly shallow

Most AI products are thin wrappers on foundation models. They don't encode deep domain understanding or real cognitive modeling.

03

The teams that publish don't ship

Academic labs optimize for citations. Product teams optimize for quarterly metrics. Nobody optimizes for both.

What We Do

Research-grade AI. Products thatship.

Most AI companies choose research or products. We do both: turning paper-only capabilities into deployable systems.

First focus: synthetic human intelligence - realistic AI populations with psychological depth, social dynamics, and evolving behavior for decision simulation.

6 products shipped. 9+ LLMs orchestrated. Peer-reviewable research.

Products Shipped
LLMs Integrated
Faster Than Surveys
Zero to Six Products

Synthetic Human Intelligence

AI systems that model how real people think, react, and make decisions. Not just pattern matching, but genuine cognitive simulation grounded in validated psychology.

Multi-Model Orchestration

We coordinate across 9+ foundation models to produce cognitively diverse, statistically valid outputs. One model's bias is another model's blind spot. We use all of them.

Production-Grade Research

Every paper we publish ships as a product. Every product we ship is backed by published research. No gap between theory and deployment. No gap between lab and production.

Research

We publish what we build.

Our research is open. We publish our methods, release our architectures, and share what we learn because the best way to advance the field is to let others build on your work.

February 2026

Layered Persona Synthesis and Adaptive Social Simulation via Large Language Models

Our first published paper introducing PersonaGen, a six-layer cognitive architecture for generating high-fidelity synthetic personas that think, remember, and evolve across simulations.

105+
Psychometric dimensions
96.5%
Coherence pass rate
6
Generation layers
9+
LLM providers
View all research

Stay in the loop

Sign up for Lab Notes by zeldaLabs

Short dispatches from the frontier of synthetic human intelligence. No spam. Unsubscribe anytime.

Products

Built from research. Shipped toreality.

Six products built on a shared foundation of synthetic human intelligence. Each one started as a research question. Each one ships today.

01 · Flagship

TownSquare AI

Collective Intelligence Infrastructure

Simulate entire communities, millions of synthetic personas that think, debate, and evolve. Powered by multi-model intelligence, our agents don't just agree. They surface hidden risks, edge cases, and the reasoning behind the results.

Explore TownSquare
1M+
Cognitive Agents
159
Parameters per Persona
9+
Foundation Models
1000x
Faster Than Surveys
02

TownFolk

AI Survey Simulation

Generate population-valid survey responses from synthetic personas. Replace months of fieldwork with hours of simulation without sacrificing statistical rigor or demographic validity.

03

TownBox

Autonomous Browser Agents

Deploy AI agents that navigate the open web autonomously. Extract data, fill forms, monitor changes, and complete multi-step workflows. No API required.

04

TownHall

Voice-First AI Debate

Multi-agent deliberation through spoken dialogue. Structured argumentation between synthetic personas that surface real disagreement and unexpected consensus.

05

ProjectKnex

AI Hiring Intelligence

Match candidates to roles using cognitive and behavioral simulation. Go beyond resumes and model how candidates actually think, collaborate, and perform.

06

Lab Infrastructure

Shared Foundation

The engine powering every product. Persona generation, multi-LLM orchestration, simulation runtime, and evaluation pipelines. When the foundation improves, everything improves.

Infrastructure

One foundation. Six products. And counting.

Every product we ship is built on the same core infrastructure. When the foundation improves, everything improves. That's the advantage of building a lab, not just a product.

Persona Generation

105+ psychometric dimensions per persona. Demographic, cognitive, and behavioral fidelity that holds up under statistical validation.

Multi-LLM Orchestration

9+ foundation models working in concert. Cognitive diversity by design, different models bring different reasoning styles to the same problem.

Simulation Runtime

Real-time simulation of millions of agents interacting, debating, and evolving. Memory systems that persist across sessions and simulations.

Evaluation Pipelines

Automated coherence testing, bias detection, and statistical validation. Every output is measured against ground truth before it ships.

Lab Partners

We work with a small number of partners.

If you're building something ambitious and want early access to frontier AI capabilities, we should talk. We're selective because depth matters more than breadth.

Early access to frontier capabilities
Direct input into our roadmap
Dedicated team that builds alongside you
Become a Lab Partner