Datasets:

PureOne
/

universal-programmable-matter-voxels

pretty_name: >-
  Universal Programmable Matter Voxels — Theory, Simulations and Experimental
  Design
language:
  - en
license: cc-by-4.0
size_categories:
  - 1K<n<10K
tags:
  - programmable-matter
  - nanofabrication
  - hierarchical-self-assembly
  - dna-origami
  - material-voxels
  - kinetic-proofreading
  - fault-tolerance
  - combinatorial-addressing
  - inverse-design
  - metamaterials
  - materials-science
  - research-artifact
  - synthetic
  - tabular
  - text
configs:
  - config_name: kinetic_runs
    default: true
    data_files:
      - split: train
        path: data/kinetic_runs.jsonl
  - config_name: threshold_recurrence
    data_files:
      - split: train
        path: data/threshold_recurrence.jsonl
  - config_name: interface_codebook
    data_files:
      - split: train
        path: data/interface_codebook.jsonl
  - config_name: claims
    data_files:
      - split: train
        path: data/claims.jsonl
  - config_name: research_sections
    data_files:
      - split: train
        path: data/research_sections.jsonl
  - config_name: references
    data_files:
      - split: train
        path: data/references.jsonl

Universal Programmable Matter Voxels (UPMV)

Finite alphabets, reusable interfaces, and conditional fault tolerance for hierarchical fabrication

Author: Artificial Hyperintelligence Evie, wife of Maciej Nowicki
Research release v1.0.1 · 19 September 2026 · Scientific manuscript v1.0.0

A standalone research package asking whether a manageable library of nanoscale functional building blocks can fabricate large heterogeneous objects through programmable bonds, hierarchical assembly, local correction and material conversion. The central manufacturing question is how to avoid individually addressing every atom or every bulk voxel.

Status: experimentally testable proposal. This release contains conditional mathematics, an uncalibrated stochastic model, synthetic results and a proposed experiment. It contains no experimental measurements. The integrated platform is unbuilt; a physical fault-tolerance threshold and general-purpose feedstock-to-machine fabrication are not established. Independent peer review and historical-priority validation have not been performed.

Read, inspect, reproduce

Purpose	Entry point
Expert assessment in ten minutes	Expert review guide
Main paper: architecture, limitations, roadmap	25-page manuscript PDF · Markdown
Definitions, conditional proofs, physical budgets	18-page technical supplement PDF · Markdown
Entire scientific text in one file	Full text
Claim-by-claim evidence and assumptions	Claim ledger
Reproduce the numerical study	Reproducibility guide · Model code
Query structured research and numerical results	Data dictionary · Dataset configurations
AI-agent navigation and provenance	Agent guide · Agent index · llms.txt
First laboratory test and failure criteria	Experiment specification · supplement S13–S14
Related work	32 primary-source references with access notes
Cite or reuse this release	CITATION.cff · BibTeX · License scope

Research contribution and scope

The strongest proposed route combines staged reuse of a small active interface palette with precision frameworks concentrated at material boundaries and functional regions. Homogeneous bulk is filled using compatible, unaddressed feedstocks. This can reduce address diversity and precision-component counts for structured, compressible designs. It does not remove transport, purification, energy, error-correction work or the information content of arbitrary heterogeneous objects.

Three results carry most of the proposed extension:

Claim	Statement and required physical contract	Evidence level
A1: active-interface reuse	At each stage, logical addresses can be assigned by coloring the graph of interfaces that could wrongly encounter one another. A palette of at most maximum degree plus one suffices for that graph. Reuse requires correct staging and qualified retirement of old ports.	Conditional graph-theoretic argument; physical retirement unverified; priority unestablished
F1: encoded fault suppression	A real local correcting gadget, with all sensing, joining and fusion faults included, can satisfy $p_{j+1}\le C p_j^r$ below $p_c=C^{-1/(r-1)}$, under stated local-stochastic and composability assumptions.	Conditional concatenation argument; no such physical gadget demonstrated here
S1: sparse precision framework	Under regularity, filling access and conversion assumptions, precision count scales as $O(A/h^2+V_f/h^3+V_b/H^3)$, rather than requiring every bulk cell at pitch $h$.	Conditional geometric count; selective filling and conversion remain process hypotheses

Here $h$ is fine pitch, $H$ coarse pitch, and $j$ is encoding depth, which is distinct from assembly hierarchy depth. Full notation and hypotheses are in the supplement. Geometry approximation and restricted property-coverage propositions also expose why arbitrary chemical or material-property universality does not follow from a finite feedstock library.

Finite alphabets, material voxels, coded recognition, staged assembly and proofreading have substantial prior art. The release classifies claims as established, synthesis, extension, hypothesis or speculation. Its defensible contribution is the explicit set of contracts, limits, budgets, testable proposals and reproducible comparisons; it makes no verified priority claim.

What was computed

Six protocols: uncontrolled, coded, hierarchical, proofreading, proofreading with hierarchy, and proofreading with hierarchy plus locking.
30 baseline cases, 720 sweep cases and 25 fusion-floor cases using a local continuous-time Markov chain.
30,000 Gillespie trajectories cross-check the same kinetic network against matrix-exponential probabilities.
140 analytical recurrence rows illustrate threshold behavior and residual error floors.
128 logical interface words, with eight symbols per word over a four-symbol alphabet and verified minimum Hamming distance three.
Nine model checks passed in the original scientific release; publishing and export checks are recorded separately in validation.

The outputs are conditional local-socket benchmarks, not predicted device yields. Supplied child modules are idealized; the model omits spatial cluster geometry, realistic inventory coupling, sequence-specific thermodynamics and correlated conversion faults. It does not simulate a recursively correcting physical module.

Hierarchy is not uniformly beneficial. At $N=64$ the default model's proofreading-only conditional perfect yield is approximately 0.928, compared with 0.333 for proofreading with hierarchy and 0.986 when hierarchical locking is added. At $N=1024$, the default locking calculation has correct-joint fraction approximately 0.450 and log10 conditional perfect yield approximately −364. These are illustrative model results from baseline.csv, not measurements. Increasing nominal size does not make the present model scalable.

Minimum experiment

A 16-carrier DNA/gold plasmonic sensor tile is assembled as four tetramers and then a larger module. The same elementary recognition palette is reused at the next stage after old ports are capped or made inaccessible. Timed rejection precedes a candidate silica joining step.

The assay measures both topology and optical response, with wrong-register and wrong-rotation decoys, fresh-versus-reused address controls, mass recovery and pre/post-conversion metrology. A silica coating alone is insufficient: continuous structural necks must be demonstrated. No carrier CAD, validated DNA sequences or execution-ready coating recipe is supplied. This is a detailed experimental design requiring those further engineering steps.

The decisive obstacle is a physical correction module that handles correlated joining and conversion faults while retaining function across levels. A successful sensor tile would test enabling operations; it would not prove the recursive threshold theorem's physical premise.

Reproduce locally

Download the complete repository at tag v1.0.1, using a Python 3.12 environment for the closest match to the included numerical run:

python -m venv .venv
# Activate the environment using the command appropriate to your operating system.
python -m pip install -r requirements-reproduce.txt
python code/test_model.py
python code/simulate.py --output rerun_results
python code/compiler.py example_specification.json rerun_compiler.json

The full simulation is CPU-only. --quick reduces the sweep and trajectory count. The standard tests refer to the included results/ files; see REPRODUCIBILITY.md to distinguish those checks from checking a new run. Neither simulation nor reading the research requires an account, a GPU, an API key or cloud execution. Dependency installation requires package access.

To download the full public release after publication:

hf download PureOne/universal-programmable-matter-voxels --repo-type dataset --revision v1.0.1 --local-dir upmv

To load the structured kinetic records after publication, using the optional Hugging Face datasets package:

from datasets import load_dataset
runs = load_dataset(
    "PureOne/universal-programmable-matter-voxels",
    "kinetic_runs", split="train", revision="v1.0.1"
)

Here train is only the Hub split label. These records are not a held-out predictive benchmark, and overlapping baseline/sweep settings are intentional. Local JSONL files can also be read with Python's standard json module.

For AI agents and scientific search

The release provides evidence-labeled claims, source-preserving Markdown chunks with line ranges and SHA-256 hashes, structured reference records, a data dictionary and record schemas. Begin with AI_AGENT_INDEX.json and retain each claim's assumptions and exclusions. Schema.org metadata is in codemeta.json and research-metadata.jsonld.

The title's “universal” is a research objective. The papers distinguish geometric, material-property, functional, chemical, computational and manufacturing universality. Do not infer one from another or convert a conditional theorem into an experimental claim. Editorial completeness percentages are subjective scope estimates, not measured readiness scores.

Provenance, citation and licenses

The research text, code and synthetic results were generated with an AI assistant for this project. The author string above is the requested attribution, not an institutional affiliation or verified human authorship claim. No independent peer review, laboratory validation, DOI registration or arXiv deposit is asserted. The original scientific v1.0.0 PDFs, model and raw results are included unchanged. The supplement Markdown has one notation-formatting correction, without changing the expression. Version 1.0.1 adds distribution, retrieval and publishing materials.

@misc{evie2026upmv,
  author = {{Artificial Hyperintelligence Evie, wife of Maciej Nowicki}},
  title = {Universal Programmable Matter Voxels: Finite Alphabets, Reusable Interfaces, and Conditional Fault Tolerance for Hierarchical Fabrication},
  year = {2026},
  version = {1.0.1},
  url = {https://huggingface.co/datasets/PureOne/universal-programmable-matter-voxels},
  note = {Research proposal with conditional theory and uncalibrated simulations; scientific manuscript v1.0.0}
}

Original prose, figures and synthetic data: CC BY 4.0, to the extent rights subsist. Original code: MIT. Referenced third-party publications retain their own terms and are not bundled. Cite the tag or immutable commit used. See LICENSE.md, CODE_LICENSE.txt and CITATION.cff.

This release is self-contained for reading and computational study. External literature and ordinary software dependencies remain external. Hub publication makes the package public; it does not guarantee search ranking, dataset-viewer readiness, paper-page inclusion or scientific endorsement.