July 2025 marked a concentrated set of advances in Artificial Intelligence applied to science, with progress spanning autonomous experimentation, computational acceleration, and multi-agent coordination. The developments described in the article show how Artificial Intelligence is accelerating hypothesis generation, speeding experiments, and enabling end-to-end research workflows that reduce time, cost, and environmental impact.
Concrete lab and platform innovations included a self-driving laboratory from North Carolina State University that collects data at rates at least ten times faster than traditional methods, lowering costs and reducing environmental impact. Stanford Medicine rolled out virtual laboratory platforms with virtual supporting scientists and principal investigators capable of performing complex biological research with minimal human input. Google’s Artificial Intelligence co-scientist, built on Gemini 2.0, expanded adoption in biomedical workflows by helping generate hypotheses and design projects, while hierarchical frameworks such as NovelSeek were presented as systems capable of managing full research cycles from literature review to experimentation and reporting. Artificial Intelligence-Scientist-v2 demonstrated independent hypothesis generation, experiment planning and execution, result analysis, and manuscript drafting, iterating outputs with a vision-language feedback loop and showing work accepted by academic workshops.
Advances in computational and genomic tools were also highlighted. DeepMind’s AlphaGenome targets the noncoding 98% of the human genome by predicting molecular features such as chromatin accessibility and gene expression to clarify how genetic variation affects function. Equivariant graph neural networks now reproduce simulations like density functional theory in under ten minutes on standard laptops, and some climate simulations reported energy consumption reductions approaching 40 percent. In drug discovery, companies including Alphabet’s Isomorphic Labs are preparing to test Artificial Intelligence-designed molecules in clinical trials, signaling a shift toward faster development paths for treatments.
Multi-agent coordination and governance issues featured as well. Manus introduced Broad Research, a platform supporting more than 100 general-purpose agents that operate independently yet cooperate toward shared objectives. The rapid pace of change prompted calls for updated infrastructure and oversight; the July policy statement America’s AI Action Plan and the 2025 AI Index Report were cited as frameworks emphasizing the need to reform experimentation approaches and build scalable systems for responsible integration of these technologies.
