On 17 September, Anthropic presented work on optimising more than 30 open models used in biomolecular research. According to the company, Claude helped improve computation speeds over a few weeks of work.
For someone outside a laboratory, the benefit may feel distant. But research software that runs faster can create room for more experiments and make computing power more useful.
These results are not the same as a finished treatment or a new medicine. They concern the tools behind the research, and the improvements depend on what is measured and the precision required. Anthropic is also sharing optimised code so others can build on the work.
It is a reminder that AI development does not only appear in our chats. Sometimes the most interesting change happens behind the scenes, where existing work becomes easier to carry out and more ideas can be tested.