The agent, called Elements Claw, completed the discovery run in roughly 28 GPU-hours, according to coverage from the South China Morning Post and Pandaily in early July 2026. The result, also reported by MSN and the Chinese AI-news outlet AIBase, frames a future in which AI systems don't just suggest candidates for scientists to test but take the initiative across the discovery pipeline. For ongoing breaking AI news on how autonomous agents are reshaping scientific research, the DAMO result is a notable data point.
An agent, not just a model
The distinction DAMO is leaning on is between a predictive model and an agent. Traditional AI-for-materials systems typically generate a list of promising compounds and hand them to human researchers for synthesis and testing. Elements Claw is described as operating more autonomously, coordinating the steps of a discovery workflow — hypothesizing, evaluating, and filtering candidates — with limited human intervention.
That framing matters because the frontier of AI in science has shifted from static prediction toward agentic systems that can plan multi-step tasks, call external tools, and refine their approach based on feedback. DAMO Academy, Alibaba's fundamental-research arm, has positioned Elements Claw as evidence that this shift can deliver real, experimentally confirmed results rather than purely computational ones.
Why superconductors
Superconducting materials — which conduct electricity with zero resistance below a critical temperature — are among the most coveted targets in materials science. They underpin technologies from MRI machines and particle accelerators to quantum computing and lossless power transmission. Finding new ones is slow and expensive through conventional trial-and-error, which is why the field has become a marquee use case for AI-assisted discovery.
The promise is straightforward: if AI can narrow an enormous space of possible compounds down to a handful worth synthesizing, it could compress research timelines from years to days. DAMO's claim that its four discoveries were confirmed by experiment — rather than remaining theoretical predictions — is the crucial caveat that separates credible AI discovery from hype.
A fast-moving field
The DAMO result arrives amid a surge of AI-driven materials discovery. In 2023, Google DeepMind's GNoME project identified roughly 2.2 million new crystal structures and roughly 380,000 stable materials, publishing the findings in Nature and releasing the database to other researchers. Since then, multiple labs have pushed toward systems that don't just predict structures but actively help design and validate them. Elements Claw represents the latest, and one of the most aggressive, attempts to put an autonomous agent at the center of that loop.
What sets the DAMO work apart, according to the reports, is the combination of speed and verification: a large volume of computational exploration compressed into roughly a day of GPU time, followed by physical confirmation that the predicted materials actually behave as expected.
The bigger picture
For Alibaba, the result is a credibility marker for its AI research ambitions at a moment when competition between Chinese and Western AI labs is intensifying. DAMO Academy has been a vehicle for the company's longer-horizon bets, and a headline-grabbing materials discovery burnishes its scientific standing even as the commercial AI market remains dominated by chatbots and coding tools.
For the broader research community, the question is whether autonomous discovery agents like Elements Claw can scale beyond a celebrated demonstration into a reliable, everyday tool. If they can, the bottleneck in materials science could shift from finding new compounds to synthesizing and testing them fast enough to keep up with what AI proposes. Either way, the idea of an AI agent that ships verified discoveries rather than just suggestions is rapidly moving from speculation to reality.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
