We extend Co‑Scientist and validate it in real‑world settings. Co‑Scientist is a Gemini‑based multi‑agent system that speeds up the full research loop—from hypothesis generation through experimental execution to manuscript writing. Moving beyond purely in‑silico hypothesis creation, the system becomes an execution‑grounded research partner, enabling closed‑loop workflows in materials science, biology and computer science.
In materials science, Co‑Scientist interfaced with a semi‑automated chemical vapor deposition (CVD) reactor to devise a safe precursor route for MXenes. The experiment produced a lamellar 2‑D material whose lattice shares key structural motifs with Ti₃C₂Tx MXene, though further work is needed to confirm the atomic structure. Leveraging Gemini 3 Deep Think, the system generated growth recipes within minutes that respect laboratory constraints, allowing a single‑attempt growth of monolayer MoS₂, MoSe₂ and WS₂ semiconductors.
In biology, Co‑Scientist predicted emergent swarming phenotypes of engineered E. coli across IPTG gradients from sparse imaging data, quantitatively matching unpublished wet‑lab morphological measurements.
In computer science, the system autonomously discovered an inference‑time scaling architecture that outperformed six frontier models on HealthBench (Hard and Professional) and reduced potential clinical harm in blinded physician evaluations.
A double‑blind study involving 30 domain experts and 450 reviews of end‑to‑end generated papers showed that Co‑Scientist’s reliability modules markedly cut hallucination and plagiarism while improving research safety.
Together, these results demonstrate the feasibility of closed‑loop multi‑agent scientific AI systems for accelerating real‑world discovery.
Blogger's Review: The study highlights the promise of tightly coupling AI with experimental platforms, yet further validation on atomic structures and larger‑scale deployments remains essential.