Inside the Virtual R&D Lab: How Human Imagination and AI Multi-Agents Shape the Future of Science

The article presents tanaike-lab, a virtual R&D lab built around LLMs, multi-agent workflows, hooks, function calling, MCP, and A2A protocols.

Using Gemini in Antigravity CLI, the author carried out an urban torrential-rain fluid dynamics study from scratch and guided it to publication.

The work ended with a manuscript posted on ESS Open Archive, showing an end-to-end AI-assisted research workflow.

It highlights how multi-agent AI tools can reduce research friction while keeping humans in charge of scientific direction.