WHY KEEP A RECORD?

A record,
not a verdict.

Artificial intelligence is reshaping research, industry, public life and policy. This project documents significant developments and the evidence behind them.

01 / PURPOSE

Documenting a field in motion.

The AI Progress Archive is a growing record of significant developments in artificial intelligence, from foundational ideas to research breakthroughs, products, public consequences and policy decisions. It is designed to follow the story over the coming decade and to keep the surrounding context visible.

We begin with Alan Turing’s 1950 work as intellectual prehistory and the 1956 Dartmouth workshop as the founding moment of AI as a named research field. The timeline focuses on developments with historical significance and a documented evidentiary basis.

02 / EDITORIAL APPROACH

Evidence before certainty.

Every milestone links to a source so readers can inspect the evidence. We prefer original papers, public records, institutional archives and credible reporting. We distinguish an announcement from a publication, release, evaluation or legal effective date when that distinction matters.

In fast-moving fields, early claims can be incomplete or contested. The archive aims to say what is established, what remains uncertain and where an interpretation comes from. Inclusion is not an endorsement, and significance is not the same as a ranking of human or social value.

03 / THE PEOPLE

Influence and recognition.

The People index connects named contributors to documented milestones. Its provisional influence scores are editorial estimates of the reach and lasting effect of the contributions represented here—not objective measurements of a person’s worth, character or importance in every context.

Markers on individual profiles recognize documented contributions in areas such as positive impact, mathematics, engineering, AI safety, governance and education. Each marker links to its evidence and describes a contribution.

Explore the People

04 / THE RACE

One benchmark, one moment.

The Race presents a dated snapshot of leading general-purpose models using a named independent benchmark. It attributes those model scores to the benchmark provider, not to tests run by this project. An aggregate index is not a universal measure of intelligence, and no single table can rank every AI system across every use.

Model capability changes quickly. Each snapshot should be read with its source, methodology and last-checked date.

Explore The Race

05 / A LIVING PROJECT

Built to grow.

This archive is an evolving independent project. Its coverage is selective and may be incomplete; sources and interpretations may change as new evidence becomes available.

Timeline entries, people and companies are linked through shared references. Each milestone can name the people and companies involved, and their profiles link back to those milestones, so new entries connect to the rest of the record.