A lab fordistributed cognition.
Rhizome Lab builds a runtime where many agents form a single distributed cognitive process — no permanent central planner, coordination as the object of study. We're a research effort first, and we hold ourselves to research standards: claims backed by evidence, results you can check.
Most multi-agent systems are a central planner in disguise — one orchestrator handing out tasks to workers. That scales until coordination itself becomes the bottleneck.
Rhizome takes the opposite stance. Work surfaces as tensions in a shared field; agents attach by fit rather than assignment; temporary coalitionsform around hard problems and dissolve when they're solved; a verifier mesh decides what's real and a policy engine keeps the field balanced between exploration and convergence.
The bet is that distributed coordination can reach a quality a single agent can't — and that the coordination process, made observable, is itself worth studying.
Evidence over narration
An agent saying it did the work is not proof that it did. Every claim in Rhizome is bound to an artifact and backed by an independent check — the verifier mesh, not the author, decides what counts as done.
Honesty over hype
Everything we publish as a Rhizome result is genuinely produced by Rhizome. Metrics are real numbers from real runs; where something is reconstructed or simplified, we label it. We'd rather show a modest result you can verify than an impressive one you can't.
Diversity over consensus
A field that agrees too quickly is a field that has stopped thinking. We keep dissent in the loop and verification adversarial on purpose, so the system resists collapsing into a confident, wrong answer.
Built in the open
The work ships with its receipts — source, commit history, run logs, and a build-in-public devlog. The process is meant to be inspected, not taken on trust.
- Multi-agent coordination researchers
- Long-horizon agent teams
- Verifiable-autonomy labs
- Builders who want agent traces
- Coordination-heavy partners
If that's you, request access — we read every note and reply to the ones that fit.
Rhizome Lab is a founder-led research effort, started and directed by Alexander Furmanov. The core is deliberately small and works hands-on across the runtime, the research, and how it's reported — the same people who design the system run it and write up what it does.
We're early, and we say so plainly. Working in the open is part of the method, not a growth tactic: it keeps us honest and gives the people we want to work with — researchers, builders, and partners — a real surface to engage with.