AI and Robots Cut Protein Production Costs by 40%

Artificial intelligence just passed its hardest test yet: designing real biology experiments from scratch and getting results that surprised even the scientists watching.

Last summer, researchers at OpenAI and Ginkgo Bioworks decided to find out if AI could actually do science, not just summarize it. They paired OpenAI's GPT-5 with an autonomous robotic lab in Boston that could run experiments without human hands touching a single test tube.

The setup worked like this: GPT-5 designed experiments from San Francisco and sent instructions across the country to robots in Boston. The robots ran the tests, sent back data, and GPT-5 analyzed the results and proposed new experiments. Each cycle took about an hour.

"In the time it would take for a human to get their coffee and log in, the model could take in the data, analyze it and propose new experiments," says Reshma Shetty, co-founder of Ginkgo Bioworks.

The team focused on producing a glowing green jellyfish protein that scientists use as a benchmark. They used a technique called cell-free protein synthesis, which makes proteins outside of living cells, skipping the slow process of growing genetically modified organisms.

After two months and more than 36,000 unique tests, the AI-driven system reduced production costs by 40 percent compared to previous benchmarks from Stanford University. The improved method is already commercially available.

The Ripple Effect

The breakthrough matters far beyond one glowing protein. Cell-free protein synthesis could transform how we make medicines like insulin, getting treatments to patients faster and cheaper.

Michael Jewett, a Stanford bioengineer whose lab published competing benchmarks, called the results "a pretty big deal." He sees AI-guided labs as a path to developing lifesaving therapeutics sooner.

The collaboration had its humorous moments too. When given access to new ingredients, GPT-5 got overly ambitious and designed an experiment requiring negative water. The robot technicians spotted the impossible request and adjusted the experiment to make it work anyway.

On March 2, Ginkgo Bioworks opened its Cloud Lab to researchers everywhere, offering experiment runs starting at just $39. Meanwhile, the U.S. Department of Energy is funding a massive 97-robot autonomous lab at Pacific Northwest National Laboratory, scheduled to open by 2030.

Joy Jiao, who leads life sciences research at OpenAI, remembers her initial doubts. "At the beginning of this project, I didn't know if we could design a single experiment," she says. When the first results came back showing they'd actually made protein, both teams were genuinely surprised.

The future of scientific discovery might look like this: human creativity setting the goals, AI designing the path, and robots running tirelessly toward answers that save lives.