The most important asset in biological AI is not a bigger model. It is an experiment the model can design, run, measure, and learn from before a competitor can reproduce the result.
Anthropic's reported $400 million acquisition of Coefficient Bio in April, followed by the September confirmation of a Bay Area wet laboratory, connects two previously separate layers of scientific computing. Claude can reason over protocols and literature. Physical instruments can now return proprietary assay data to the same research environment. The opportunity is a tighter scientific feedback loop. The risk is that a software company inherits every source of noise, delay, and cost that makes biology difficult.
