Kevin Weil, head of OpenAI for Science, pushes that analogy himself. “I believe 2026 might be for AI and science what 2025 was for AI in software program engineering,” he stated at a press briefing yesterday. “We’re beginning to see that very same sort of inflection.”
OpenAI claims that round 1.3 million scientists around the globe submit greater than 8 million queries every week to ChatGPT on superior matters in science and math. “That tells us that AI is transferring from curiosity to core workflow for scientists,” Weil stated.
Prism is a response to that person conduct. It may also be seen as a bid to lock in additional scientists to OpenAI’s merchandise in a market filled with rival chatbots.
“I largely use GPT-5 for writing code,” says Roland Dunbrack, a professor of biology on the Fox Chase Most cancers Heart in Philadelphia, who isn’t linked to OpenAI. “Sometimes, I ask LLMs a scientific query, mainly hoping it will probably discover data within the literature quicker than I can. It used to hallucinate references however doesn’t appear to do this very a lot anymore.”
Nikita Zhivotovskiy, a statistician on the College of California, Berkeley, says GPT-5 has already grow to be an vital device in his work. “It typically helps polish the textual content of papers, catching mathematical typos or bugs, and offers typically helpful suggestions,” he says. “This can be very useful for fast summarization of analysis articles, making interplay with the scientific literature smoother.”
By combining a chatbot with an on a regular basis piece of software program, Prism follows a pattern set by merchandise reminiscent of OpenAI’s Atlas, which embeds ChatGPT in an internet browser, in addition to LLM-powered workplace instruments from companies reminiscent of Microsoft and Google DeepMind.
Prism incorporates GPT-5.2, the corporate’s greatest mannequin but for mathematical and scientific problem-solving, into an editor for writing paperwork in LaTeX, a standard coding language that scientists use for formatting scientific papers.
A ChatGPT chat field sits on the backside of the display, under a view of the article being written. Scientists can name on ChatGPT for something they need. It could actually assist them draft the textual content, summarize associated articles, handle their citations, flip pictures of whiteboard scribbles into equations or diagrams, or speak by hypotheses or mathematical proofs.
