AQCat Adsorption Spin Now Accessible Via Waitlist Integration

SandboxAQ is now offering limited access to AQCat Adsorption Spin, its spin-aware machine learning engine for heterogeneous catalysis, and researchers must join a waitlist to utilize the tool. This development indicates considerable demand for an AI designed to accelerate the complex process of catalyst discovery, starting with the critical adsorption energy calculation. Unlike many platforms that rely on existing experimental data, SandboxAQ generates its own physics-grounded training data, a key advantage when tackling areas where public resources are limited. The company’s approach extends beyond isolated modeling; its three-layer infrastructure aims to deliver complete workflows and accelerate scientific breakthroughs from concept to physical testing.

Physics-Grounded Data Generation for Large Quantitative Models

SandboxAQ is challenging conventional AI development in materials science and drug discovery by prioritizing data generation rooted in physics, rather than relying on existing experimental datasets. This approach addresses a critical limitation of many current platforms, which struggle when publicly available data is sparse. Unlike competitors, SandboxAQ’s Large Quantitative Models (LQMs) actively create training data using high-fidelity physics engines like Density Functional Theory (DFT) and Molecular Dynamics (MD), simulating molecular behavior to reflect actual physical laws. This allows researchers to progress even when empirical data is lacking, avoiding bottlenecks that often halt progress in areas with limited prior research. The company’s three-layer infrastructure is central to this capability; the first layer focuses entirely on generating this physics-grounded data, guided by the specific problem rather than available methods. This synthetic data is then used to train proprietary AI and physics models, forming the second layer.

AQCat, a model designed for heterogeneous catalyst screening, achieves performance at near-DFT accuracy, 20,000 times faster than traditional DFT methods. Another model, AQAffinity, predicts protein-ligand binding affinity approximately 1,000 times faster than physics-based FEP simulations, without requiring a crystal structure. Importantly, SandboxAQ isn’t solely focused on proprietary development; they are also committed to collaborating with the Open Source community and improving public models, exemplified by AQAffinity being built on and enhancing OpenFold3. However, the LQMs extend beyond simply providing models; the third layer automates workflows to accelerate the entire discovery process, allowing scientists to concentrate on critical decisions rather than routine execution.

AQCat Adsorption Spin Integration with Large Language Models

SandboxAQ is extending the reach of its Large Quantitative Models (LQMs) by integrating them with leading Large Language Models (LLMs), offering researchers a new avenue for accelerating scientific discovery. This convergence allows access to complex AI through familiar natural language prompting, streamlining the process from initial hypothesis to tangible results. The company’s first integration, AQCat Adsorption Spin, is currently available via a waitlist, indicating considerable demand for this specific tool designed for heterogeneous catalysis, a crucial process in many industrial applications. This data generation begins with the scientific problem itself, rather than being constrained by available methods, enabling targeted analysis of specific conditions and chemistries. The company states that they “generate our own physics-grounded training data — and decide which data to generate by starting from the problem, not the method,” highlighting a key advantage over competitors.

The resulting data fuels proprietary AI and physics models, including AQCat, which can screen heterogeneous catalyst candidates at near-DFT accuracy, but at a speed 20,000 times faster than traditional DFT methods. According to SandboxAQ, “Our agents allow researchers to shift their focus from executing each step to making the decisions that matter,” and the platform supports applications across diverse fields including drug discovery, solid-state battery materials, and alloy design, all within a shared infrastructure and continuously evolving dataset. Researchers can now access AQCat Adsorption Spin through Claude, enabling them to prioritize the most promising catalyst candidates before committing significant resources to modeling and laboratory evaluation.