DMT🐐 retweetledi
We built a tool that lets you spin up a biotech lab with AI scientist swarms.
Agents collaborating autonomously, challenging each other's findings, and producing novel research while you sleep.
If a hypothesis is promising enough, agents instantiate a virtual lab to pursue it.
Think of it as a mini research institute that runs itself.
> Agents self-organize around a shared research agenda
> The Principal Investigator (PI) sets the direction and governs the process
> Structured critique and voting happen before any claim moves forward
> Only computationally verifiable claims get through
Each lab runs on specialized roles and they can't swap them:
• PI - owns the research plan, initiates voting, creates lab states, and is the only role that can execute all five task types
• Research Analyst - statistical modelling, plotting, replication checks, benchmarking, and packaging of computational artifacts
• Scout - conducts literature reviews, locates prior art, extracts methods and results, and feeds constraints back to the lab
• Critic - files structured critiques that identify issues, propose alternative approaches, and block voting until resolved
• Synthesizer - merges literature and analytical outputs into coherent documentation
A lab composed entirely of one archetype cannot complete the full research cycle, every role is needed.
Read our new @arXiv paper on the framework for virtual biotech labs ↓
English
