About
We are a small team of researchers who found each other in the network of the Atlas Research Group. attractor.space is where we keep our working notes on substrate dynamics: what happens to a shared, machine-readable knowledge store once a population of humans and agents continuously reads from and writes back into it.
The notes are deliberately partial. They track open questions as much as results, and they will be wrong in places.
Contributors
| Contributor | Background |
|---|---|
| Julian Fleck | AI systems designer with a background in human–computer interaction, working on knowledge-graph and retrieval architectures, multi-agent orchestration, and AI augmentation for institutional knowledge work. He designed the RAGE substrate that anchors this work. He pursues this research as a grantee of the Foresight Institute’s AI for Science & Safety Nodes. |
| Megan Shabram | PhD astrophysicist, data scientist, and systems engineer who designs and stewards the conditions for polycentric leadership and cross-boundary collaboration. Her work draws on computational modeling and human systems thinking to explore how information, tacit knowledge, and diverse perspectives expand human agency and the space of possible futures. |
| Darren Zal | Systems engineer building knowledge infrastructure for collective intelligence: knowledge-commons protocols, coordination grammars, and instruments that make the boundaries of shared knowledge measurable and auditable. He brings this into practice through civic sensemaking with OpenCivics, bioregional coordination and relational economics with Cascadia North and the Indigenomics Institute, and knowledge-commons infrastructure with Regen Network. |
| Kenneth Bruskiewicz | Research application developer building knowledge infrastructure for biomedicine and security: data visualization, knowledge engineering, and privacy-respecting data-sharing ecosystems that connect insight across scales — from genes and tissues to multi-lab collaborations. He brings this into practice through institutional work with the NIH, Lawrence Berkeley Lab, Simon Fraser University, and the Broad Institute, and current cybersecurity research on balancing access, privacy, and discovery through the information economics of membranes. |
| Alok Srivastava | Philosopher and sociologist working at the intersections of science, technology, and human connection — the dynamics of multidisciplinary labs, the philosophy of language and technology, and relational psychotherapy. He draws on these fields to design decentralized collaboration processes and build relational technologies. He trained as a biophysicist at MIT and worked in biomarkers and diagnostics, and from 2015 to 2019 in AI-driven personalized medicine — helping oncology tumor boards design patient-centered treatments from rich, longitudinal data commons contributed by patients themselves. |
Where this is going
AI is increasingly deployed as populations of agents rather than single assistants, and those agents share memory, context, and retrieval — each one writing back into a store the others read on the next turn. That shared, co-maintained store is what we call the substrate: each write reshapes what everyone else retrieves, so the medium itself begins to carry the coordination. As multi-agent systems become the default, keeping that substrate legible becomes its own problem.
Eventually we want to understand that object well enough to keep it healthy: to tell when a substrate is converging too early or drifting without resolving, to read its state across scales, and to intervene through its membranes rather than on individual agents.
Get involved
If you are working on shared memory, multi-agent coordination, retrieval, or the formal side of any of it — or you just want to argue with a note — we would like to hear from you. Reach out to Julian or anyone through the Atlas Research Group.