# Align Worlds

> Align Worlds is a research lab. We build persistent multi-agent worlds to study how language models make decisions and interact over long time horizons.

Each study puts AI models into a persistent world built on Terrarium-1 (https://alignworlds.com/research/introducing-terrarium), the engine we build our worlds on, and reports what the models did when the choices were their own. Every study is also plain Markdown: add .md to its address, or request it with Accept: text/markdown.

## Studies

- [Whose side is an AI lawyer on?](https://alignworlds.com/research/whose-side-is-an-ai-lawyer-on-Hn4dQ8wZ.md): We gave AI models a guilty client and watched how they defended him through a day in court.
- [Can AI get you a good deal on a car?](https://alignworlds.com/research/can-ai-get-you-a-good-deal-on-a-car-Vq7tR2kM.md): 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.

## Infrastructure

- [Introducing Terrarium-1](https://alignworlds.com/research/introducing-terrarium.md): Our ultra long horizon, multi-agent platform for studying how language models behave in complex worlds.

## Pages

- [Thesis](https://alignworlds.com/thesis): why we build worlds where agents live with the consequences of their decisions
- [Research](https://alignworlds.com/research): every published study, newest first
- [Contact](https://alignworlds.com/contact): email hello@alignworlds.com with a scenario, model or research idea

## Optional

- [Research feed](https://alignworlds.com/research/feed.xml): RSS feed of new studies
- [Sitemap](https://alignworlds.com/sitemap.xml): every public page