Large language models (LLMs) are increasingly deployed as agents for scientific discovery, but standardized frameworks for evaluating their performance and behavior in scientific workflows are lacking...

Biological AI learns primarily from existing observations, but many questions cannot be answered from available data alone. Here we show that AI can instead acquire knowledge by acting directly on biological systems and learning from the consequences. We developed a closed-loop framework in which autonomous agents design protein variants, construct and characterize them in a robotic laboratory, learn from the resulting experimental feedback, and decide what experiments to perform next. We then allowed the system to operate continuously and without human intervention for approximately one month, during which multiple agents independently explored protein sequence space while learning from shared experimental experience. Applied to glycoside hydrolases, the agents discovered enzymes with substantially altered substrate specificity toward non-native sugars and progressively learned the structure of the underlying sequence-function landscape. The resulting experimental experience also revealed determinants of substrate specificity and protein expression that were not specified as learning objectives. These results demonstrate that AI can autonomously interact with biology over extended periods to acquire knowledge through experience, establishing a framework for biological discovery driven by continuous experimental interaction. ### Competing Interest Statement The authors have declared no competing interest. National Institute of General Medical Sciences, 5R01GM150929

Antibodies originate from germline templates and are diversified by somatic hypermutation, producing sequences in which conserved germline residues scaffold structure while rare non-germline (NGL) sub...

Mutational effect transfer learning (METL) is a protein language model framework that unites machine learning and biophysical modeling. Transformer-based neural networks are pretrained on biophysical simulation data to capture fundamental relationships between protein sequence, structure and energetics.

To better understand and design proteins, it is crucial to consider the multifunctional landscapes on which all proteins exist. Proteins are often optimized for single functions during design and engi...

Protein language models (pLMs) enable generative design of novel protein sequences but remain fundamentally misaligned with protein engineering goals, as they lack explicit understanding of function a...

Advances in metagenomics, deep learning, and generative protein design have enabled broad in silico exploration of sequence space, but experimental characterization is still constrained by the cost an...