Co-Scientist: A multi-agent AI partner to accelerate research

2026-05-12 · Source: Google DeepMind News · Field: Science & Research — Life Sciences & Biology, Research Methodology & Innovation, Health & Medical Research · Depth: Advanced, long

Summary

Co-Scientist, a new multi-agent AI system built with Gemini, was published in Nature on May 19, 2026, designed to accelerate scientific research by generating and refining novel hypotheses. Developed by Google DeepMind, Google Research, Google Cloud, and Google Labs, this system is now available to individual researchers via the "Hypothesis Generation" experimental tool. Co-Scientist operates through three phases: generating initial ideas, debating their validity and novelty with virtual peer reviewers and ranking agents, and iteratively evolving the most promising hypotheses. It integrates web search and specialized databases like ChEMBL and UniProt, and can leverage models such as AlphaFold for verification. The system has been validated in life sciences, contributing to breakthroughs in areas like liver fibrosis, ALS, cellular aging, and infectious diseases, with an enterprise version also being previewed.

Key takeaway

For AI Scientists and Research Scientists seeking to accelerate discovery, Co-Scientist offers a powerful multi-agent AI partner for hypothesis generation and refinement. You should explore integrating this Gemini-based system to overcome information overload and rapidly identify novel research directions, potentially reducing experimental lead times from months to days. Remember that Co-Scientist is a collaborative tool, requiring your scientific expertise to validate and guide its outputs, ensuring ethical and impactful research outcomes.

Key insights

Co-Scientist is a multi-agent AI system that iteratively generates, debates, and refines scientific hypotheses to accelerate discovery.

Principles

• Scientific discovery benefits from structured, iterative ideation and critique.
• Multi-agent AI systems can mimic diverse research team roles.
• Rigorous hypothesis verification is crucial for AI-generated scientific claims.

Method

Co-Scientist employs specialized Gemini-based agents for idea generation, debate (critique, ranking via "idea tournament" with pairwise comparisons), and evolution, orchestrated by an adaptive supervisor agent. It verifies claims against literature and databases.

In practice

• Register for Hypothesis Generation to access Co-Scientist.
• Utilize AI for initial hypothesis generation in complex problems.
• Apply multi-agent systems for structured scientific reasoning.

Topics

• Multi-agent AI Systems
• Hypothesis Generation
• Scientific Discovery
• Gemini Model
• Life Sciences Research
• AI in Drug Discovery

Best for: AI Scientist, Research Scientist, Director of AI/ML

Related on AIssential

• Accelerating scientific discovery with Co-Scientist — Co-Scientist, a Gemini-based multi-agent AI, accelerates scientific discovery by generating and refining novel, verifiable hypotheses.
• Meet the AI "Co-Scientist" Changing Everything 🤖🧪 #ai — Co-scientist is a fine-tuned Gemini with tools for scientific hypothesis generation, data analysis, and literature summarization.
• Generating novel scientific hypotheses with Co-Scientist — Co-Scientist is a multi-agent AI system designed to accelerate scientific discovery by automating literature review and hypothesis generation.
• 4 ways researchers are collaborating with Co-Scientist to solve big problems — Co-Scientist is a multi-agent AI system accelerating scientific hypothesis generation and refinement across life sciences.
• Gemini for Science: AI experiments and tools for a new era of discovery — Gemini for Science leverages AI agents to automate complex research tasks, significantly accelerating scientific discovery and exploration.
• 12 AI Co-Scientists of 2026 — AI co-scientists are rapidly accelerating scientific discovery across diverse fields through autonomous hypothesis generation, experimentation, and analysis.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Google DeepMind News.