On Tuesday, OpenAI launched Prism, a novel scientific workspace application freely available to anyone with a ChatGPT account. Positioned as an AI-enhanced word processor and research instrument for scientific papers, Prism is deeply integrated with GPT-5.2, allowing users to assess claims, refine prose, or search for existing research.
Prism is not designed to conduct research autonomously or without human supervision. Executives anticipate that it will accelerate the efforts of human scientists, drawing parallels between Prism and coding interfaces such as Cursor and Windsurf.
“I believe 2026 will be for AI and science what 2025 was for AI and software engineering,” remarked Kevin Weill, VP of OpenAI for Science, during a press call announcing the tool.
This new web-based software arrives as OpenAI observes a substantial influx of scientific queries directed at its consumer products, including ChatGPT. The company reports that ChatGPT processes an average of 8.4 million messages weekly on advanced topics in the hard sciences — though the exact number originating from professional researchers is hard to ascertain.
AI-assisted research is also gaining prevalence among academic researchers. In mathematics, AI models have been employed to prove several long-standing Erdos problems through a combination of literature review and novel applications of existing techniques. While the implications of these proofs remain a topic of intense discussion, the results represent an early success for advocates of AI models and formal verification systems.
A statistics paper published in December utilized GPT 5.2 Pro to establish new proofs for a central axiom of statistical theory, with human researchers solely providing prompts and verifying the model’s output. OpenAI commended this achievement in a blog post, presenting it as a blueprint for future human-AI collaboration in research.
“In domains with axiomatic theoretical foundations,” the post articulated, “frontier models can aid in exploring proofs, testing hypotheses, and identifying connections that would otherwise require substantial human effort to uncover.”
Much of the value in OpenAI’s new system stems from straightforward product development applied to established standards. Prism integrates with LaTeX, an open-source system commonly used for formatting and typesetting scientific papers, yet it significantly expands upon most available LaTeX software tools. The program additionally leverages GPT 5.2’s visual capabilities, allowing researchers to compile diagrams from online whiteboard drawings, addressing a notable difficulty with current tools.
Perhaps its most potent feature derives from merging the typical capabilities of an AI model with more stringent context management. When users access a ChatGPT window through Prism, the model gains access to the full context of the research project, resulting in responses that are both more pertinent and more intelligent.
While a proficient GPT-5.2 user could accomplish much of this, OpenAI hopes that a more intuitive interface will more quickly attract scientific researchers. Weill characterized it as the same confluence of factors that made AI tools so effective in software engineering.
“Software engineering accelerated in part because of amazing models,” he informed reporters, “and in part because of deep workflow integration.”