The CBR Open Experimental Network is an open, distributed collaboration investigating whether dimensionality, topology and physical behaviour can arise from minimal computational systems — with no spatial structure written in as a prior.
Minimal relational rules, with no spatial coordinates.
Persistent structure appears as a web of connections.
Connectivity begins to behave like a manifold.
Distance, dimension and curvature become measurable.
Known physical behaviour recovered — quantitatively.
Computation → Relations → Topology → Geometry → Physics · each arrow is a hypothesis that can fail on its own.
Participation never requires accepting CBR as a theory of reality. You might join because you think it's promising, because you think it's unlikely, or because the mathematics is interesting on its own terms. Every one of those is an equally legitimate reason — the network is built around cooperative disagreement.
Can computational systems with no predefined spatial geometry generate persistent structures whose large-scale behaviour exhibits dimensionality, manifold topology, geometry and eventually physical dynamics?
The question decomposes into progressively stronger propositions. Each must be independently challengeable — and failure at any rung is a scientifically meaningful result.
Failure is a finding. If the ladder breaks at H3, that tells us something real about computation and geometry.
Before execution, wherever possible, three roles agree on what would count. This keeps an experiment from being defined only by the people who expect a particular answer.
Explains the case for the hypothesis and what a supporting result would look like — the constructive reading of the evidence.
Defines competing explanations, artefacts and the failure criteria that would sink the claim — before any data exists.
Determines only whether the agreed test was actually satisfied — independent of who hoped for which outcome.
A successful criticism is not an obstacle to the network. It is a contribution to it.
The programme runs from the first testable question up to quantitative physics. Each experiment carries a persistent identifier and a version, and each is designed to be able to fail independently of the others.
Running the same code is not the same as independently reproducing an experiment. CBR-OEN distinguishes three increasingly strong forms — strong claims normally require the third.
The same implementation and the same seed reproduce the result. The floor: it confirms the pipeline is deterministic and intact.
An independent implementation — built from the specification, not the reference code — reproduces equivalent results.
Multiple independent institutions reproduce the finding. The most significant form of agreement the network recognises.
REQUIRED FOR STRONG CLAIMSA preregistered specification is frozen and hashed before results exist. Every run then emits a hash-addressed manifest — so provenance travels with the evidence, not alongside it.
experiment: CBR-E001 version: 1.0 title: Emergent dimensionality from relational computation hypothesis: # effective dimension may emerge from: geometry-free computation null_hypothesis: No reproducible dimensional attractor occurs. initial_conditions: geometry_encoded: false spatial_coordinates: false excluded_objectives: - spatial_dimension - spectral_dimension - manifold_score randomness: deterministic_from_seed: true analysis: preregistered: true
{
"experiment": "CBR-E004:v1.0",
"implementation": "JULIA-02",
"institution": "NODE-A17",
"seed": 428972114,
"code_hash": "9f2c…a1",
"specification_hash": "c7e0…4d",
"environment_hash": "11ab…9f",
"input_hash": "6d3e…22",
"result_hash": "be91…0c",
"started": "2026-06-04T08:12Z",
"completed": "2026-06-04T09:47Z",
"status": "complete"
}
# a run is not an output file —
# it is a hash of everything
# that produced the output.Institutions and researchers undertake one or more defined roles. A hypothesis group never decides whether its own hypothesis succeeded — separation of roles is designed in.
State the claim, the null, and its falsification conditions.
Define or review the dimension, homology and manifold tests.
Build to the spec in any language — independent code is encouraged.
Run registered batches on university, HPC or independent compute.
Perform statistics — often blinded to which rule or code produced the data.
Repeat an experiment independently, with no hand in the original build.
Hunt for artefacts, hidden assumptions and alternative explanations.
CBR-Lab provides reference results — but the spec, not the code, is authoritative.
No department needs to endorse an unconventional theory to take part. It only needs to do the thing it is already good at.
Review the definition of a manifold or homology test.
Build an independent implementation from the specification.
Execute registered run batches across a seed range.
Review the preregistration and the inference procedure.
Assess correspondence with known physical theory.
Develop alternative minimal computational rule families.
Attempt an independent replication of a published finding.
Attempt to falsify the experiment. A good challenge counts.
CBR-OEN succeeds not if CBR is supported, but if it produces precise hypotheses, reproducible simulations, rigorous tests, independent implementations, honest replications and transparent negative results. If CBR fails, the network should tell us why. If it survives, the network should tell us how strongly.
Participation does not imply endorsement of Compute-Based Reality or any associated interpretation.