May 19, 2026
Peer reviewed, comment roasted
Era: From Nature publication to catalyzing Computational Discovery
Google’s science AI just hit Nature—and the internet is split between awe and side-eye
TLDR: Google says its new AI tool ERA can help scientists write and test research code well enough to support real projects like disease and water forecasting. Online, people are torn between calling it AI’s most useful glow-up yet and warning that science doesn’t need a faster way to be confidently wrong.
Google rolled out ERA, an artificial intelligence tool that helps scientists write and improve research software, and the official pitch is huge: faster discoveries, smarter forecasts, and less time lost wrestling with code. It’s already tied to Google’s new Computational Discovery tester program, and Google says it performed at expert level across jobs like disease forecasting, water prediction, and tracking carbon dioxide in the atmosphere. In plain English: the company wants an AI lab assistant that can help researchers test ideas much faster.
But the real fireworks are in the crowd reaction. One camp is basically yelling, “This is the good version of AI!” These commenters are thrilled that the tech isn’t just making memes or spammy slop, but helping with flu forecasts, water planning, and climate research. The other camp is hitting the brakes hard, asking whether “expert-level” means truly reliable or just very confident autocomplete in a lab coat. The hottest argument? Whether this democratizes science for smaller teams—or just gives Big Tech an even bigger lead over universities already struggling for money.
And yes, the jokes arrived instantly. People compared ERA to a PhD student who never sleeps, a science intern powered by coffee and GPUs, and “the one lab partner who does all the work but might also accidentally delete the spreadsheet.” In short: excitement, suspicion, and plenty of gallows humor about the future of research getting a shiny new robot co-author.
Key Points
- •Google says its AI system Empirical Research Assistance (ERA) was published in Nature and helped build the Computational Discovery prototype now available to trusted testers in Google Labs.
- •ERA is described as a Gemini-based scientific coding tool that can search literature, write and optimize code, test alternatives, and evaluate outputs using a tree-search approach.
- •Google reports that ERA achieved expert-level performance on benchmark tasks across genomics, public health, satellite imagery analysis, neuroscience prediction, time-series forecasting, and mathematics.
- •The article says Google Research and collaborators have produced eight manuscripts applying ERA to specific scientific problems, including five newly released papers.
- •Applications highlighted include U.S. respiratory disease hospital-admission forecasting, California spring runoff prediction outperforming Bulletin 120, and high-resolution atmospheric CO2 mapping from geostationary satellite data.