Related Experiment Video
Updated: May 21, 2026

Discovery and Synthesis Optimization of Isoreticular Al(III) Phosphonate-Based Metal-Organic Framework Compounds Using High-Throughput Methods
Published on: October 6, 2023
Accelerating scientific discovery with Co-Scientist
Juraj Gottweis1, Wei-Hung Weng2, Alexander Daryin3
1Google Cloud AI Research, Zurich, Switzerland. juro@google.com.
Co-Scientist, an AI system, aids scientists in generating and refining novel hypotheses for faster discovery. This artificial intelligence tool accelerates research through structured thinking and experimental validation in biomedicine.
Area of Science:
- Artificial Intelligence
- Biomedical Research
- Scientific Discovery
Background:
- Scientific discovery relies on hypothesis generation and experimental validation.
- Augmenting human scientists with AI can accelerate the discovery process.
- Novel AI systems are needed to assist in formulating and refining research hypotheses.
Purpose of the Study:
- Introduce Co-Scientist, a multi-agent AI system for structured scientific thinking and hypothesis generation.
- To assist scientists in discovering new knowledge by formulating novel hypotheses for experimental verification.
- Demonstrate the system's potential in biomedical applications, including drug repurposing and target discovery.
Main Methods:
- Developed a multi-agent AI system (Co-Scientist) using Gemini.
- Implemented an asynchronous task execution framework for flexible compute scaling.
- Utilized a tournament evolution process for self-improving hypothesis generation and refinement.
- Focused validation on drug repurposing, novel target discovery, and antimicrobial resistance mechanisms.
Main Results:
- Co-Scientist successfully generated and refined novel hypotheses.
- Automated evaluations demonstrated improved hypothesis quality with increased test-time compute.
- Identified new drug repurposing candidates and synergistic therapies for acute myeloid leukemia.
- In vitro experiments validated the identified therapeutic strategies.
Conclusions:
- Co-Scientist accelerates scientific discovery by empowering AI-assisted scientists.
- The multi-agent architecture and evolutionary process enhance hypothesis generation.
- Real-world biomedical applications show the system's potential to expedite research and clinical translation.
Related Concept Videos
Tandem Mass Spectrometry
The Scientific Method
The Scientific Method
Generally, predictions are tested using carefully-designed experiments. Based on the outcome of these...
The Scientific Method
Drug Discovery: Overview
Cofactors and Coenzymes

