Discovered Materials offers $1 million for five unsolved problems to measure AI progress

Five open materials-science problems now have a $1 million total prize pool attached. Discovered Materials is offering the money to establish concrete measures of AI’s scientific progress—not to celebrate a breakthrough already achieved. The startup says frontier models worked autonomously on the problems for days and made little headway.

The challenge, reported by TechCrunch on November 5, seeks theories and numerical methods rather than one new substance. The company recruited an academic advisory group to help define the five problems.

Days of work, little headway

Before announcing the prizes, Discovered Materials gave frontier AI agents internet access and other resources to work independently for several days on each problem. Their limited progress is the company’s account, not an independently verified performance result.

Co-founder Advaith Sridhar wants the challenges to track whether models are improving in materials science. The identity of the eventual solver is secondary: he told TechCrunch that solutions from his company, a professor or a frontier AI lab would all serve the purpose of deepening scientific understanding.

Two theories, three computational methods

• Explain cuprate superconductivity. The challenge seeks a theory explaining how these materials can exhibit zero electrical resistance at very low temperatures.
• Predict silicon–silicon dioxide interfaces. A numerical method would determine the structures where these materials meet—a combination found in almost all semiconductors.
• Connect grain structure with heat behavior. A numerical method would link a material’s grain structure to its thermal properties, enabling better heat-dissipation materials.
• Explain changes caused by light. The requested theory would predict why shining light on a material changes its behavior, particularly around superconductivity.
• Improve Density Functional Theory. The challenge seeks a better numerical method for this technique, used to model the structures of atoms and molecules.

Setting the questions before seeking answers

Stanford professor Felipe Jornada helped develop the challenges. In a statement, he said explaining observed but unexplained phenomena would improve understanding of physics and shape electronics, energy storage and manufacturing.

Sridhar describes the approach as establishing the challenges and verification first, then inviting explanatory ideas. He credits the professors with helping identify the right questions.

The challenge sits alongside Discovered Materials’ own semiconductor research. The startup develops materials for chip applications, works with chip companies on research projects and funds scientific efforts of its own.