AI boosts wet lab cloning efficiency by 79x

OpenAI tested GPT-5’s ability to improve real-world biology—not just crunch theory. Working with Red Queen Bio, they set up a cloning experiment where the model proposed tweaks to the protocol, scientists ran the tests, and GPT-5 iterated based on results. The outcome? A 79-fold improvement in cloning efficiency. The secret sauce was a novel mechanism using two enzymes, RecA and T4 gene 32 protein, which the AI suggested. This matters because cloning is the bread and butter of genetic engineering—think protein design, gene screens, and strain building. If AI can optimize mundane lab steps like this, it could free up researchers to focus on bigger questions. But there’s a catch: the work was done under tight biosecurity controls, highlighting both the promise and the risk of letting AI tinker with biology. Why it matters: This experiment shows AI moving from pure reasoning to hands-on lab work, proving it can be a practical co-pilot in scientific discovery—a shift that could dramatically accelerate biotech innovation while demanding smarter safeguards.