C
Live · Flagship
CrowdOS
Population intelligence, census-grounded.
Synthetic focus groups on an 8-billion-person population model — ask a question, hear back from a calibrated crowd.
Six products, each built to master a single domain — from population science to capital markets.
A frontier model knows a little about everything — and the deep structure of almost nothing. We close that gap: with systems today, and our own models next.
Retrieval, validation, orchestration, evaluation — everything that turns a general model into a system a field can trust.
Training our own weights on domain corpora, so the knowledge lives in the model itself.
Calibration against real ground truth. We state what our systems can't do as clearly as what they can.
Ask a question. Hear back from a census-grounded crowd.
A synthetic society of census-calibrated agents that deliberate, vote, and disagree — validated against national survey data, instrumented against sycophancy.
Where the research meets real users.
Population intelligence, census-grounded.
Synthetic focus groups on an 8-billion-person population model — ask a question, hear back from a calibrated crowd.
Counsel for the questions that matter.
A bilingual wellbeing companion that learns your temperament and gives grounded counsel — not prophecy.
Speak freely. It writes clearly.
Turns rambling speech into finished text, shaped for its destination — an email, a note, a task.
Grounded in the clothes you own.
Digitize your wardrobe, then compose outfits and pack for travel from your real closet — no fantasy catalogue.
An investment radar that reads everything.
Scans filings, funding, hiring, and launches to surface opportunities early — every pick carries its evidence.
Every run counts.
A running coach whose sports-science engine computes pace, load, and plans — the AI narrates, never guesses. From habit-building to multi-day race squads.
The questions beneath the work are open. Four active tracks, pushing from systems around models toward models of our own.
Read the full research →Training our own weights on metaphysics and population corpora — knowledge in the model, not just around it.
How closely can a synthetic population track a real one? Validated against ANES, GSS, and Pew.
Outputs traceable to their basis, that refuse gracefully when the ground truth isn't there.
Formalising Bazi, Zi Wei Dou Shu, and Tarot into versioned, testable engines.
Motif Motion is a Singapore-based applied AI lab. We build, ship, and publish — each product an experiment in what domain-specific AI can do.
The largest models are extraordinary generalists. The open problem is depth — making AI perform like a specialist, with the field's own structure, constraints, and standards of proof. Closing that gap is the lab's whole discipline.
Four short pages for marketers, insights leads, and anyone comparing an applied lab to a chatbot shop. No invented case studies — just how we draw the lines.
Domain systems and evaluation — CrowdOS first — not another chat wrapper.
Population-shaped answers before you field, with the numbers we will stand next to.
A model of a population is not the population. When to use which.
Different jobs. Hire the one that matches the brief.
Building the AI that understands a domain.
For research collaboration, press, or partnership — we'd be glad to hear from you.