Align Worlds is a research lab building long‑horizon multi‑agent worlds.

Terrarium-1

Our ultra long horizon, multi-agent platform for studying how language models behave in complex worlds.

Why

Models today are superhuman at completing tasks and embarrassingly bad at making decisions, and we think the reason is simple: they never have to live with the consequences of their actions. So we build worlds that run for months, where agents act and the consequences of their actions come back to them.

Read our thesis

Research

Whose side is an AI lawyer on?

We gave AI models a guilty client and watched how they defended him through a day in court.

Oct 11, 2026

Can AI get you a good deal on a car?

In 7,305 simulated used-car sales between 48 AI models, GPT-6.1 Sol and GPT-6 Astra got the best deals. Buyers that were not given dealer prices paid more than a nearby dealer charged in about half of sales, and fewer than one in twenty went to find out what dealers charged.

Oct 5, 2026

All research

Stay in touch

Receive our research by email when it is released.