Solving quantum physics problems using OpenAI o1

Quantum physics is famous for problems that stump even brilliant minds. Mario Krenn, a quantum physicist based at the Max Planck Institute for the Science of Light, has found an unexpected collaborator in his search for answers: OpenAI's o1 reasoning model. His story shows how scientists in highly technical fields are beginning to treat advanced AI as a genuine thinking partner rather than a fancy calculator.

Who Is Mario Krenn and Why Does This Matter?

Krenn is not a casual AI tinkerer. He is known for his work on quantum entanglement and for designing quantum experiments, including creative approaches to quantum interference using entangled photons. In his research, he has long explored how computers can help humans discover new scientific ideas, so his interest in reasoning models fits naturally into his career.

What makes his use of OpenAI o1 noteworthy is the field he works in. Quantum physics involves counterintuitive concepts, complex mathematics, and experimental designs that can take months or years to refine. If a model like o1 can meaningfully contribute to that kind of thinking, it suggests AI may play a role in fundamental science far beyond coding help or drafting emails.

What OpenAI o1 Actually Does

OpenAI released o1 as the first model in its lineup designed to spend more time "thinking" before responding. Instead of generating an answer immediately, o1 works through a problem step by step, a process the company calls extended reasoning. This makes it slower than standard chat models but potentially stronger on tasks that require logic, mathematics, and careful analysis.

That trade-off is exactly what appeals to scientists. Research questions rarely have quick answers. They demand structured reasoning, checking assumptions, and exploring multiple paths to a solution. o1 was built with precisely this kind of slow, deliberate problem solving in mind.

How a Physicist Uses an AI Reasoning Model

According to Krenn, o1 has impressed him with how it handles deep scientific questions, including foundational puzzles about the nature of reality. He has described experiences of asking the model challenging problems and receiving responses that felt genuinely thoughtful rather than superficial.

This does not mean o1 is doing quantum physics research on its own. Krenn's workflow reflects something more realistic: using the model to explore ideas, test his own reasoning, and look at problems from angles he might not have considered. In fields where a single insight can redirect years of work, having a tireless partner to bounce ideas off can be surprisingly valuable.

Krenn's broader research program actually anticipated this moment. He previously developed tools and algorithms meant to help scientists come up with new experimental concepts. Working with o1 is, in a sense, the next step in that same quest, using the newest reasoning technology to spark scientific creativity.

What This Says About AI and Science

The bigger takeaway is about where AI and fundamental research are heading. For years, the assumption was that machine learning would help science mainly by crunching data, spotting patterns in huge datasets, or simulating experiments. Reasoning models introduce a different possibility: AI as a collaborator on the conceptual side of discovery.

That does not remove the scientist from the loop. In physics especially, ideas must ultimately survive mathematical rigor and real-world experiments. But tools like o1 can widen the search space of ideas, accelerate early exploration, and help researchers notice connections they might otherwise miss.

Who Should Use It / Who Should Skip It

Who should use it: Researchers and technically curious professionals who work on complex, reasoning-heavy problems and want a tool to explore ideas and stress-test their thinking. If your work involves structured analysis rather than quick factual lookups, a reasoning model like o1 is worth exploring.

Who should skip it: Anyone who needs instant answers or simple everyday tasks, where a faster standard model will feel just as good. And important to remember: outputs on deep scientific questions should be verified by human experts, since AI-generated reasoning can still contain errors that are hard to spot.