A small foundation working on a big question.

We started from a simple worry and an equally simple hope. The worry is that AI is growing capable much faster than our public institutions are adapting. Without clear ground rules, ordinary people could lose both the ability to see what software is doing in their name and any real say in how it works. The hope is that, with the right guardrails, the same tools can help communities think better together than any of us could alone.

When cars first shared town streets with people on foot, towns responded with stop signs, licence plates and driving tests. Software that acts for people needs the equivalent rules, and few of them exist yet.

Our work is research and civic engineering for the next set of roads. We ask how a world of capable software agents stays understandable and accountable to the people living in it. We write our proposals in plain language, test them in simulation code, and publish everything openly so you can inspect our reasoning, improve what works, or prove us wrong.

Why an independent foundation?

Most writing about AI today comes from two places. Either it comes from companies selling a product, who naturally want you to believe their software is inevitable and harmless. Or it comes from distant commentators who point out what is broken without building a practical alternative you can actually test.

We wanted a third kind of workshop: more like a municipal water laboratory or a consumer testing kitchen. In a testing kitchen, nobody is paid a commission to praise a toaster. The engineers plug the appliance in, measure whether the cord overheats, write down the results in plain words, and publish the blueprint for a safer plug so any manufacturer can use it.

What we are, and aren't

How the pieces of our site fit together

If you are visiting for the first time, here is the shortest map of what is on our workbench:

• Vision explains our five everyday guardrails: a separate road for agents, an assistant loyal to you, power on loan, a shared notebook, and public computing infrastructure.
• Research separates what is already proven in peer-reviewed science from what we propose and what remains unsolved, alongside our open standards drafts.
• Labs houses our interactive tools, including Shared Horizon (a scenario engine and 348-organisation guidance atlas) and the Observatory (where we track real-world indicators and score our own forecasts).
• Blog translates technical questions, from medical billing audits and pricing bots to capability leases and community rule-mending, into plain-language essays.

Much of this is unsolved, and we say where. The open problems page lists what we haven't worked out and what would help.