
Introducing DiG and the Discovery Community
A new benchmark for discovery, and a community built around exploring it.
We started Thinking About Thinking because a lot of important science happens when curious people get together to work on hard questions.
Here's what we're doing next.
An independent group of researchers, including Thinking About Thinking founders Dr Ruairidh Battleday and Professor James Whittington, has released DiG: Discovery in Games, new research exploring how discovery can be measured in humans and frontier AI systems.
Alongside the research, the group has developed DIG.bench: a benchmark designed to test whether an AI system can work out an unknown rule through experimentation.
DIG.bench doesn't ask a system to reproduce something it may have seen before. It puts humans and machines into unfamiliar interactive environments. To succeed, you have to experiment, form hypotheses, learn from failure and work out the underlying rules.
Alongside DiG is DisCo, the Discovery Community. It's for researchers, scientists, engineers, students, mathematicians and strong problem solvers interested in discovery, reasoning and intelligence.
Thousands of people have already joined to play the discovery worlds, test their problem solving and compete on the leaderboard.
Thinking About Thinking has joined DisCo as a Founding Community Partner.
For the Fellowship, DisCo is somewhere to carry on the conversations from our seminars and events, through shared experiments, scientific puzzles and solving problems together.
After the first DisCo competition, applications for Cohort 002 are now open.
Cohort 002 has new environments, a new competition and a new leaderboard.
If you're a Thinking About Thinking Ambassador, you're already in cohort 001. To join as a community captain, email clare@thinkingaboutthinking.org
Whether you work in machine learning, neuroscience, mathematics, cognitive science, physics, engineering or something else, DisCo tests how you handle problems nobody has explained to you.
You explore a world, form a hypothesis, test it, get it wrong and try again until you work out the rule.
It's also a community built around one of the main questions behind Thinking About Thinking:
How do humans and machines discover?
Fellows and Ambassadors
The people who run our work in their own institutions and communities.
Read more →Our Partnerships
Our partnerships with universities, societies and public-interest organisations.
Read more →AE Global Summit
Three days on the open problems in AI, 25 to 27 November 2026.
Read more →