OpenAI's experimental reasoning model wins gold at the IMO!

Dear Readers,

How does it feel when machines suddenly think like humans – only faster, more persistent, and more self-critical? OpenAI has developed a model that doesn't respond in seconds, but spends hours deliberating, finding its own errors in reasoning, and writing mathematical proofs that even IMO gold medalists recognize. What we are witnessing is more than an AI update – it could be the first real step toward universal thinking.

In this issue, we dive into this groundbreaking moment for AI research and show how AI is transforming our markets, central banks, and financial analysis. You'll learn why the Fed is betting on AI but deliberately saying no to it when it comes to interest rate decisions, and how Anthropic is ushering in a new era for analysts with Claude. Curious about how it all fits together? Then stay tuned – it's worth it.

In Today’s Issue:

All the best,

In a dramatic shift in AI research, OpenAI has unveiled an experimental reasoning LLM that solved five out of six International Mathematical Olympiad (IMO) problems under real competition conditions – achieving gold level.

What makes this model so groundbreaking is not only the number of correct solutions, but above all the way it gets there: instead of responding in seconds, the AI thinks for hours, checking its own chains of reasoning and refining its arguments in multiple loops – just like a human mathematician. This approach combines “long thinking” with self-verification processes and a strategic search for plausible solution paths in natural language. The result is multi-page, complex proofs that have been recognized even by former IMO gold medalists.

Unlike earlier systems such as AlphaGeometry, which were highly formalized, OpenAI is pursuing a generalist approach with this model. The goal is not to solve a specific class of tasks, but to develop universal reasoning – a step toward true artificial general intelligence. For research, this could be the beginning of an era in which AI assistants not only understand proofs, but also generate new scientific insights.

Why it matters: This model shows for the first time that LLMs can not only reproduce logic, but can also develop creative, consistent lines of reasoning over hours. This opens up potential for AI applications in physics, chemistry, and other disciplines that were previously considered “too creative.”

Sources:

🔗 https://www.businessinsider.com/openai-gold-iom-math-competition-2025-7

🔗 https://www.lesswrong.com/posts/RcBqeJ8GHM2LygQK3/openai-claims-imo-gold-medal

AI won't take over the world. People who know how to use AI will.

Here's how to stay ahead with AI:

Join 1 million pros and start learning AI

A new interactive reasoning benchmark, ARC-AGI-3, has been released with a $10K contest to challenge AI agents on tasks that are easy for humans but currently impossible for even the most advanced AI models.

U.S. Federal Reserve Governor Lisa Cook reports that AI (especially LLMs) is used extensively to support research, text creation, and early warning systems – but deliberately not in the monetary policy decision-making process. This distinction demonstrates a structural understanding: AI boosts productivity and can reduce inflationary pressures in the medium term, but it can also drive investment. For investors and economists, this means that monetary policy will be prepared more efficiently but will remain human-driven – which is important for market stability and transparency.

Anthropic is introducing the first domain-specific AI platform for financial data – Claude integrates various sources such as PitchBook, Morningstar, and Daloopa. This innovation marks a turning point for analysts: a unified interface that centralizes data aggregation, analysis, and reporting – pure efficiency. Business models could shift: platform providers have two options – cooperation with AI providers such as Anthropic or in-house development. For investors, this means margin potential, but also consolidation risks in the data market.

An article in Barron's discusses how AI could reduce inflation in the medium term by lowering costs (e.g., automation) and increasing efficiency. At the same time, it points out that AI-based productivity gains could raise neutralized interest rates – and thus not diminish the importance of central banks. This two-phase effect – relief versus structural interest rate shifts – represents a potential upheaval for macroeconomics and portfolio valuations. Decision-makers must pay attention to how AI markets shape dislodged inflation and interest rate norms.

Working on the future of finance through AI? From trading algorithms to credit modeling and fintech infrastructure, we’re looking for work that explores the intersection of AI and financial systems.

Submit your paper or project to Superintelligence, the top AI newsletter with 200k+ readers, by emailing [email protected] with the subject line “Finance Submission”. We’ll contact you if we’d like to feature it.