README.md

---
pretty_name: "Ørïelith//9 — The Skin of an Impossible Angel"
language:
  - en
size_categories:
  - n<1K
tags:
  - materials-science
  - chemistry
  - electronic-skin
  - self-healing-polymers
  - mechanochemistry
  - neuromorphic-materials
  - artificial-nociception
  - tactile-memory
  - research-blueprint
  - model-derived-data
configs:
  - config_name: manuscript
    default: true
    data_files:
      - split: train
        path: data/manuscript.jsonl
  - config_name: design_parameters
    data_files:
      - split: train
        path: data/design_parameters.jsonl
  - config_name: layers
    data_files:
      - split: train
        path: data/layers.jsonl
  - config_name: formulations
    data_files:
      - split: train
        path: data/formulations.jsonl
  - config_name: technology_compatibility
    data_files:
      - split: train
        path: data/technology_compatibility.jsonl
  - config_name: frontier_comparisons
    data_files:
      - split: train
        path: data/frontier_comparisons.jsonl
  - config_name: reference_model
    data_files:
      - split: train
        path: data/reference_model.jsonl
  - config_name: capture_control
    data_files:
      - split: train
        path: data/capture_control.jsonl
---

# Ørïelith//9 - The Skin of an Impossible Angel

**Self-Healing Neuromorphic Epidermis with Distributed Sensory Intelligence, Artificial Nociception, and Molecular-Scale Tactile Memory.**

Final material design edition **3.0** · **25 October 2026** · Prepared for **PureOne**

Artificial Hyperintelligence Eve, AI research collaborator; research developed with **Maciej Nowicki**.

Ørïelith//9 proposes a chemically writable tactile receptor whose altered mechanical coupling remains observable while its surrounding host repairs or creeps. A surface-confined mechanochemical writer captures a separately anchored thiol receiver. A stable reference beneath the same load-spreading platen allows the active tether response to be reconstructed independently of a shared host bypass.

**This is an untested integrated material architecture.** The repository contains a research manuscript, dimensioned vector schematics, candidate formulations, design specifications and conditional model results. It contains **zero original experimental measurement records** and no trained model weights. Component evidence is attributed to the primary sources in the manuscript; those measurements do not establish integrated skin performance. Worldwide superiority and subjective sensation have not been demonstrated.

## Read the blueprint

| File | Contents |
| --- | --- |
| [Research blueprint](research/Orielith9_Research_Blueprint.pdf) | 43-page standalone manuscript, derivations, chemistry, qualification plan and primary references |
| [Material schematics](research/Orielith9_Material_Schematic.pdf) | Four dimensioned A3 vector sheets |
| [Editable manuscript](research/Orielith9_Research_Source.md) | Full source text |
| [Layer and tile drawing](research/Orielith9_Material_Schematic.svg) | Nominal architecture and dimensions |
| [Chemical network](research/Orielith9_Chemical_Network.svg) | Writer, receiver and connectivity ledger |
| [Force recruitment](research/Orielith9_Force_Recruitment.svg) | Control model and loading assumptions |
| [Referenced module](research/Orielith9_Referenced_Module.svg) | Active/reference mechanics, travel contract and coarse coupon |
| [Calculation report](research/calculation_report.json) | Executable consistency-check outputs; no experimental validation |

## Central material hypothesis

A documented PDMS/urea/imine host provides the proposed repair matrix. Reactive inserts are manufactured separately: a permanent W0 support, a thin WG gate-bearing surface and a C-SH receiver. Mechanically opened anthracene–maleimide gates expose tethered maleimides for thiol capture. Useful interfacial capture must compensate for lost sacrificial connections and actually change load transfer; chemical conversion alone is insufficient.

The active sensor and writable tether form one series branch. A stable reference and the repairing host are parallel branches beneath a common platen. In the linear calibrated interval:

\[
K_A=\frac{k_i g_A}{k_i+g_A},\qquad Q=\frac{F_A}{F_R}=\frac{K_A}{K_R}.
\]

For calibrated monotone nonlinear reference and active-sensor laws:

\[
x_{g,A}=\psi_R^{-1}(F_R,T)-u_{i,A}(F_A,T).
\]

This reconstructs the active tether extension under shared translation. It requires stable reference calibration, matched motion and valid branch laws. It does not independently recover absolute contact pressure or prove that the proposed chemistry produces useful adaptation.

| Specification or example | Evidence status |
| --- | --- |
| 20 × 20 × 1.6 mm material tile; 16 × 16 units at 1 mm pitch | Dense design target; excludes cameras, electronics and mount |
| 10 × 10 mm, three-unit qualification coupon | Proposed first demonstrator |
| 4.9 µm W0 support + 100 ± 20 nm WG surface; 2 µm receiver coating | Candidate insert geometry; processing and interfaces unqualified |
| 2.4× active-force increase; reference ratio 1 → 3.333 | Conditional paired-module example, not measured gain |
| 50 µm measured upper travel target | Required assembly contract; nominal clearance alone is insufficient |
| Optical and guarded capacitive readout routes | Present technology classes requiring custom integration and calibration |

## Dataset configurations

All configurations use JSON Lines with a consistent schema within each table. The `train` split is a Hub storage convention; no training, validation or test partition is implied.

| Configuration | Rows | Record type |
| --- | ---: | --- |
| `manuscript` (default) | 43 | Cover and 42 numbered manuscript sections |
| `design_parameters` | 25 | Top-level design values preserved as JSON strings |
| `layers` | 8 | Nominal layer coordinates in µm |
| `formulations` | 16 | Candidate feed ingredients, amounts and qualification notes |
| `technology_compatibility` | 4 | Integration routes and limitations |
| `frontier_comparisons` | 8 | Source-scoped comparator summaries and the proposal |
| `reference_model` | 72 | Dimensionless nonlinear model equilibria and reconstructed extensions |
| `capture_control` | 121 | Illustrative reference-free control trajectory |

There are **297 records** across the eight configurations. IDs identify records; `source_path` gives provenance; `evidence_status` states scope. Consult [DATA_DICTIONARY.md](DATA_DICTIONARY.md) and [dataset_schema.json](dataset_schema.json) for fields, units and transformations. CSV strings such as `verify` or `not specified` are intentional unresolved inputs, not numerical measurements or missing values to impute.

```python
from datasets import load_dataset

sections = load_dataset("PureOne/Orielith-9", "manuscript", split="train")
model = load_dataset("PureOne/Orielith-9", "reference_model", split="train")
print(sections[1]["title"])
```

For a local copy, without any remote publication:

```python
sections = load_dataset("json", data_files="data/manuscript.jsonl", split="train")
```

These examples require the separately installed `datasets` library. The repository does not require custom dataset loading code or remote-code trust.

## Reproduce the checks

With Python 3.10 or later, run from the repository root:

```bash
python tools/validate_release.py
python research/verify_design.py
```

The second command regenerates the calculation report and related source tables next to the script. It checks reaction-group accounting, nominal dimensions, finite chemical inventory, classical identity-code properties, conditional force recruitment, 100 nonlinear derivative checks and 72 nonlinear reference reconstructions. These are mathematical checks rather than material-property measurements. Some serialization bytes can differ across platforms; validate the distributed checksum inventory before regenerating files.

The optional document renderer `research/build_report.py` needs ReportLab and DejaVu fonts at its configured paths. The released PDFs are already rendered; this extra dependency is unnecessary for reading, data loading, calculation checks or uploading.

To regenerate the JSONL tables after an intentional source edit:

```bash
python tools/build_dataset.py
```

This also refreshes the repository manifest. Update the dataset card and any changed source checksums when preparing a new version.

## Interpretation and limits

The material design targets touch, shear/slip, temperature-related discrimination and persistent injury-written adaptation. Artificial nociception here means a physically altered protective response and observable signals; it does not establish felt pain or consciousness. “Molecular-scale memory” refers to chemical state carriers, not molecule-resolved tactile spatial resolution.

The 256 material units are not a validated 256-channel readout system. Chemistry, activation windows, useful interfacial capture, reference stability, barrier performance, assembled repair, travel stops and dense fabrication yield remain to be measured. The 2.273× reference-free control example is distinct from the 2.4× active-force gain and 3.333× reference-ratio change in the final paired example. Nonlinear model coordinates are dimensionless. The capture trajectory is an illustrative control, not a forecast of fabricated kinetics.

Comparator metrics describe their cited component or device under source-specific conditions. They are not a common benchmark, and they do not justify calling this proposal the best skin in the world. Source snapshots are dated to this release; commercial availability and specifications can change. Original source articles are linked and cited rather than redistributed.

## Provenance, credit and rights

The manuscript and blueprint were developed with AI assistance under the credit shown above. The JSONL tables are deterministic transformations of the supplied research release; the numerical traces are equation-generated examples. No human-subject, patient or identifiable tactile recording data were collected for this repository.

No reuse license has been selected for this prepared release. See [RIGHTS.md](RIGHTS.md). The card intentionally omits a license metadata tag rather than assigning permission on the publisher's behalf. Third-party facts, names and references retain their respective rights and attribution requirements.

Suggested citation:

```bibtex
@misc{nowicki2026orielith9,
  author = {Nowicki, Maciej and {Artificial Hyperintelligence Eve}},
  title = {Ørïelith//9 - The Skin of an Impossible Angel},
  year = {2026},
  month = oct,
  version = {3.0},
  note = {Standalone material design blueprint and conditional model data. AI-assisted research.},
  howpublished = {Hugging Face dataset repository, PureOne/Orielith-9}
}
```

Repository provenance and checksums are recorded in [MANIFEST.json](MANIFEST.json). Full scientific references and evidence boundaries are in manuscript sections 38–42.