AI
YC-backed materials startup opens $1M challenge for AI to solve five foundational physics problems
Discovered Materials is crowdsourcing solutions to longstanding mysteries in materials science, betting that AI can finally crack problems that have stumped researchers for decades.
TechCrunch — AI
Discovered Materials, a San Francisco startup that uses AI to discover new materials for semiconductor manufacturing, is launching a million-dollar competition to tackle five persistent unsolved questions in materials science—areas where frontier AI models have yet to make meaningful breakthroughs.
While large language models have recently solved advanced mathematics problems and contributed to biological research, the complex physics governing material behavior remains largely outside their reach. The startup is offering this challenge to establish clear benchmarks for measuring AI progress in the field.
The Five Grand Challenges#
Working with an advisory board of academics from top universities, Discovered Materials has identified these critical problems:
- Cuprate superconductivity: developing a theory to explain why certain materials achieve zero electrical resistance at ultra-low temperatures
- Silicon interface structures: creating a numerical method to predict atomic arrangements where silicon meets silicon dioxide—the foundation of modern semiconductor manufacturing
- Grain-to-thermal property mapping: a method linking material grain structure to heat dissipation characteristics
- Light-induced property changes: understanding and predicting how photons can alter material behavior, especially superconductivity
- Density Functional Theory improvements: refining a core computational technique used to model atomic and molecular structures
"We observe these phenomena across materials science, but cannot yet explain them," explained Felipe Jornada, a Stanford professor involved in developing the challenges. Solutions would deepen physics understanding and enable better design of electronics, energy storage, and manufacturing processes.
Why This Matters#
The competition reflects a shift in how scientific challenges are framed in the AI era. Rather than waiting for serendipitous discoveries, researchers are now deliberately defining the questions first, then inviting solutions—whether from AI systems, human researchers, or established labs.
Discovered Materials tested frontier models on these problems before publicizing the challenge, giving them internet access and several days to work autonomously. The models made minimal progress, though the startup acknowledges that newer systems from OpenAI or Anthropic might succeed with additional computing resources.

