About Field Notes

Field Notes collects what people are learning about AI and the environment. It covers two kinds of work: projects that use AI to monitor, forecast, protect or restore the environment, and work that shrinks AI's own footprint, such as the energy and water data centers use.

Every entry is built around learnings. A learning is one thing a team tried and what happened. Failures and surprises count as much as successes, because the next team can save months by knowing what didn't work.

The site is written for nonprofits and funders in plain language. Researchers and technical teams are welcome too, and can filter by machine-learning technique.

Where entries come from

Teams share their own projects. We also draft some entries from public sources so the collection is useful from the start. Those entries carry a red "unclaimed" warning until someone from the project's organization claims them, checks them, and adds what they learned.

Anyone can suggest an edit. Every change is reviewed before it goes live, and the change history on each entry shows who made it.

Acknowledgments

The idea grew out of conversations on the Walking Softer Greece 2026 trip.

• Kostas described a climate images project where AI made the team's data usable. It showed that AI can do concrete good, beyond simply doing less harm.
• Dimitri argued that failures matter as much as successes, and that teams should be braver about sharing them. That shaped the learnings at the heart of this site.
• Jonathan described a project on data-center water use, which brought tracking and open-data projects into scope.

Keeping it light

The site is open source and runs on a small budget. It uses AI only once per submission, to draft the form from a document, and a small model does that job.