World models in 2026: places you can walk through, lives still to come
World models now generate explorable places that hold together for minutes. The next frontier is behaviour over years, many lives at once, and results checked against a reality the model has never seen.
When a model speaks for a population
Language models answering as real people, scored once on survey answers nobody had read: a frontier model met 0 of 6, prompted voices overstated births threefold, and carrying the last shares forward beat a small model. Lives fitted to records came closer, a model's estimate counted as a prior meant fewer people to ask, and subgroup splits landed inside the survey's own noise in the United States and in 30 of 31 African countries.
Twenty years ahead: forecasting how long people live, on years nobody had read
Living Worlds' life expectancy forecasts, frozen twenty years back and scored once on years no run had seen: met for women and for men on the latest national figures, not met against 2024 national figures or across 155 countries, with honest intervals for women and a stacked UN model that did not help.
Synthetic populations: where simulated people hold up, and where they fail
Census-built populations, agents grounded in interviews and language models speaking for whole groups have all advanced in 2025-26. The evidence now shows clearly which of them match real people, and the dividing line is what each simulated person is built from.
Simulation for physical AI: the robot is well modelled, the people around it less so
Physical AI now runs on open physics engines, world foundation models and synthetic demonstrations measured in hundreds of thousands. The newest benchmarks ask whether simulated scores predict real ones, and the part of the scene least checked against reality is the people.
Simulation first: power through feeders nobody had measured, and comfort in fifteen countries
A feeder model that learned from Living Worlds' checked simulation first, then from a handful of real measurements, predicted 30 unmeasured low-voltage feeders better than the same model on real measurements alone; power through 40 unmetered feeders beat physics and a learned model; thermal comfort in fifteen countries came close but was not met.
Scored once: benchmark contamination and the case for pre-registered tests
In 2025-26 a public leaderboard, a flagship coding benchmark and several careful audits showed how quickly a test stops measuring once it has been seen. The remedy is old and simple: fix the test before looking, hold the data out, score it once and publish the miss.
Rooms never heard: predicting sound and light before they are measured
Living Worlds predicted how real rooms sound in arrangements and positions nobody had measured, and how daylight falls on office desks on days nobody had read: met for a room never seen (20 of 20), four rooms never seen (4 of 4), sound beyond what was measured and daylight at the desk; not met in a second room.
Hearing a room before it is built
Spatial audio got a full international standard in 2025, wave-based simulation reached open datasets, and learned acoustic fields now predict how a room sounds from a handful of measurements. The open question is how far those predictions hold beyond what was measured, and how to show it.
Families from lived records: children and siblings, counted before the answers were read
A population lived month by month in the United States from 1950 matched how many children Americans have had and how many brothers and sisters they report, closer than a frontier model answering from what it has read. The father's side, first counted through partnerships, missed, and became the next test.
Emulating the Earth: what AI weather teaches simulation of lives
Learned weather models are now operational, faster and often more skilful than physics on ordinary days, and weaker on record-breaking ones. The habits that made them trustworthy, daily scoring against what happened and honesty about extremes, are what simulation of lives and economies needs next.
Digital twins of cities and grids: what a twin needs before it is trusted
Europe's grid operators, its cities and its climate programme all moved digital twins from plans to infrastructure in 2025-26. The architecture is now well specified; the step that decides whether a twin can be relied on is a prediction scored against what it never measured.
Crowds on days nobody opened
Living Worlds' crowd against real pedestrians in Jülich experiments, a shopping centre in Osaka and six recorded street scenes, each scored once on recordings nobody had opened: met in the shopping centre, in four fresh Jülich experiments and with a learned walking layer; not met in the first Jülich set and in three of six street scenes.
Agents that remember, and people around a working machine
An agent that remembers what a person said predicted their answers four years later better than a persona, in every panel and on every question; an agent given only a life record did no better than a persona. And Living Worlds' crowd predicted how close real people came to a real robot in 7 of 8 recordings nobody had opened.
A billion lives on one GPU, and what only shows at scale
Living Worlds runs more than a billion life-months a second on one desktop GPU and on one desktop processor, with every birth, death, partnership and separation identical across engines; a million-life world that anyone can replay; and an effect of who is born that appears only when lives have to find each other.