Seeing the map of the globe light up as researchers worldwide accessed the protein structures we released was a humbling and amazing experience.

We couldn't have achieved this without the foundational work of the structural biology community, which provided the initial 150,000 protein structures that our system learned from.

In just a few years, our small team was able to determine the structures of 200 million proteins, compared to the 150,000 structures revealed by tens of thousands of scientists over 50 years.

There's enormous potential in specializing general techniques for specific applications, such as protein structure prediction, and we are actively exploring this in our AI for science work.

The launch of the AlphaFold server is crucial for providing the scientific community with access to our models, facilitating non-commercial academic research.

AlphaFold 3 is a significant step towards solving the structure prediction problem, but we still aim for 100% accuracy in understanding molecular interactions.

The tools we develop become useful when they lower the barrier to adoption, similar to how tools like BLAST and AlphaFold are widely used in biology.

AlphaFold 3 represents a significant breakthrough in understanding biomolecular structure, allowing for experimental-level accuracy in predicting protein structures and their interactions with small molecules, DNA, and RNA.

AlphaFold has become a standard tool for biology and medical research, with over 2.5 million researchers worldwide using it in their critical work, significantly impacting the scientific community.