Mycelium Fosters Human-AI Team Science with Shared Context Graphs

Sutanay Choudhury, Jeffrey J. Czajka, Lummy M. O. Monteiro, Erin Bredeweg, Jason McDermott, Katherine Wolf, Alex Beliaev, Josh Elmore, Paul Piehowski, Kylee Tate, Yuqian Gao, Aivett Bilbao, Kelly Stratton, Scott Baker, Jaydeep P. Bardhan, Kristin Burnum Johnson, Chris Oehmen, Robert Rallo· July 16, 2026 View original

Key takeaways

  • Complex scientific problems benefit from "networked intelligence" between humans and AI.
  • Mycelium is a shared workspace that automatically connects researchers and AI agents.
  • It captures, tracks, and routes observations and hypotheses to relevant team members.
  • Empirical tests show it improves cross-expert collaboration and experimental design.

Who benefits

PharmaceuticalsBiotechnologyMaterials ScienceAerospaceAcademia

Summary

This paper introduces Mycelium, an active shared workspace designed to cultivate "networked intelligence" by connecting human researchers and AI agents as a multi-user co-scientist. It automatically captures and routes observations and hypotheses to inform team decisions.

Current AI-for-science systems often focus on enhancing individual reasoning processes, but complex scientific problems are typically solved by diverse teams. The concept of "networked intelligence" aims to scale connections between humans and AI, ensuring that insights generated in one context can inform others. Mycelium is presented as an active shared workspace that embodies this principle. Mycelium functions as a multi-user co-scientist, automatically capturing critical observations and hypotheses as humans and AI agents collaborate. It tracks the relationships of these insights to the team's evolving model and intelligently routes them to the most relevant person, agent, or instrument for action. An empirical test in a biological multi-omics campaign demonstrated Mycelium's effectiveness, where routed shared context transformed a local analytical finding into a cross-expert mechanistic constraint, ultimately leading to an improved experimental design. The system provides a computational framework for sparse conditional computation over distributed scientific contexts, highlighting when a networked approach is superior to standalone agents.

Why it matters

Professionals in R&D, scientific research, and complex problem-solving domains can leverage networked intelligence to enhance collaboration between human experts and AI, accelerating discovery and improving decision-making in multi-disciplinary teams.

How to implement this in your domain

  1. 1Identify collaborative projects where human-AI interaction is critical but fragmented.
  2. 2Explore integrating shared context graph technologies like Mycelium to centralize observations and hypotheses.
  3. 3Design workflows where AI agents can automatically capture and route relevant information to human experts.
  4. 4Pilot a project using this approach to evaluate its impact on cross-functional communication and decision quality.
  5. 5Develop metrics to measure the efficiency and effectiveness of networked intelligence in specific scientific or R&D tasks.

Original post by Sutanay Choudhury, Jeffrey J. Czajka, Lummy M. O. Monteiro, Erin Bredeweg, Jason McDermott, Katherine Wolf, Alex Beliaev, Josh Elmore, Paul Piehowski, Kylee Tate, Yuqian Gao, Aivett Bilbao, Kelly Stratton, Scott Baker, Jaydeep P. Bardhan, Kristin Burnum Johnson, Chris Oehmen, Robert Rallo

"arXiv:2607.13220v1 Announce Type: new Abstract: Most AI-for-science systems focus on scaling a single reasoning process through better models, larger context windows, long-horizon agentic execution, or digital co-scientists working with one principal user. However, challenging sc…"

View on X

Originally posted by Sutanay Choudhury, Jeffrey J. Czajka, Lummy M. O. Monteiro, Erin Bredeweg, Jason McDermott, Katherine Wolf, Alex Beliaev, Josh Elmore, Paul Piehowski, Kylee Tate, Yuqian Gao, Aivett Bilbao, Kelly Stratton, Scott Baker, Jaydeep P. Bardhan, Kristin Burnum Johnson, Chris Oehmen, Robert Rallo on X · view source

Want to go deeper?

Turn these trends into skills with Learnijoy's hands-on AI & tech courses.

Explore courses