Scientific neighbors
Fractalish is not claiming a vacuum. Neighboring fields supply essential tools and warnings. The current audit statement is modest: we have not identified another public program combining these exact layers.
Boundary: neighboring work does not prove Fractalish, and Fractalish does not absorb neighboring work as a component. Each relationship is labeled as conceptual, implemented, proposed, or external reported result.
Computational epistemic complexity, persistent state, and governed memory
Marc Finzi et al., arXiv:2601.03220v2. Contributes the epiplexity framing: learnable structure versus residual surprise under bounded observer and model class.
Yanbo Zhang and Michael Levin, arXiv:2607.18433v1. Contributes a reservoir-based closed-form estimator or approximation of epiplexity with a fixed bounded observer.
Yanbo Zhang essay / conceptual article. It frames complexity around the observer and motivates the finite-observer discussion.
Ding, Nannapaneni, Liu, and Zhang, arXiv:2606.30306. Surveys persistent memory, state, and governance in LLM agents.
A Survey on Long-Term Memory Security in LLM Agents: Attacks, Defenses, and Governance Across the Memory Lifecycle, arXiv:2604.16548.
Contributes morphogenetic memory, bioelectric regulation, and diverse intelligence context. Fractalish overlaps conceptually on form, history, and control, but does not claim physical CNT memory or biological proof.
Contributes observer/action loops and model-relative behavior. Fractalish differs by emphasizing explicit receipts, replay, target contracts, and host authority boundaries.
Contributes rigorous state reconstruction and observer-relative structure. Fractalish overlaps conceptually, but adds governance, receipt custody, and public claim mapping.
Contribute memory architectures and retrieval practice. Fractalish focuses on provenance, contradiction, target-relative weighting, and host-owned action.
Contribute auditability, immutable event logs, and reconstruction. Fractalish uses these as engineering neighbors for receipts and rollback.
Contributes embodiment and structure-as-computation. Fractalish keeps morphology claims partial, target-relative, and non-unique.
Epiplexity separates structure learnable by a bounded observer from surprise that remains unlearnable under that observer and model class. Ageometrics asks what a chosen representation failed to preserve relative to a declared target, protocol, and fuller admissible record. The concerns are neighboring, but the mathematical objects and loss contracts are different.
Epiplexity's residual and Ageometrics' NGR both resist treating all information as equally usable, but they measure different losses under different contracts.
| Dimension | Epiplexity / learnable novelty | Ageometrics / NGR |
|---|---|---|
| Question asked | What structure can a bounded observer extract and reuse? | What did a declared representation fail to preserve for a target? |
| Observer model | Bounded observer/model class. | Declared observer, representation, protocol, and comparator. |
| Target dependence | Conditioned by observer/model objective. | Explicit target-contract dependence. |
| Residual definition | Unlearnable surprise under the observer/model class. | Target-relative performance gap relative to fuller admissible record. |
| History dependence | Depends on what the observer can learn from data. | Depends on what history the representation preserves or erases. |
| Persistence | Metric or reward signal by itself. | Representation audit; persistence requires separate architecture. |
| Governance | Not a governance layer by itself. | Can feed HOLD and specificity posture, but does not decide action alone. |
| Current evidence status | External reported result in this register. | Specification/working paper plus local Specificity evidence where separately claimed. |