The founding release of Nuclear Compute: a model of computation in which a radiation field and the nuclear states it drives are the processor, the memory, and the interconnect, and the decay of the same medium is the power supply.
This is not a software release in the ordinary sense. It is the first complete edition of a research program: a falsifiable theory, its criterion evaluated against the complete nuclear data record, a reference device specified to assembly grade, and seven executable instruments that recompute every number in the repository from public data in about one minute of CI time.
What this edition contains
- The theory. The foundational document (README) states the model, the gate set, the physical limits, and the honest complexity ceiling, and concentrates the entire feasibility question into one falsifiable component: the keystone gate, judged by three inequalities (leak, amplification, rate). The supplement (theory/THEORY.md) carries the engineering form of the criterion, level restoration, the Bernstein universality theorem (function approximation without the keystone), the degree homogeneity theorem (correctness never decays, only speed), the boundary theory, the nuclear lamp, and the rate coded canon.
- The criterion, run against nature. /gates screens all 1870 isomers in NUBASE2020 and evaluates the candidates against ENSDF, with an experiment menu of required cross sections per facility and a computed interconnect channel plan.
- The machine, simulated decay by decay. /simulator runs the Tier 1 sampler against exact enumeration and proves out the remaining components: adder, clock, bistable cell, telemetry link. /transport verifies every routine gate to Monte Carlo precision and includes the degree checker, a design rule tool.
- The device. /transistor is the reference transistor at three scales with one pinout, its datasheet and characteristic curves, and six accompanying notes: the sealed self sustaining ampoule (computed), its assembly grade build specification, the γ ray valve, the full component inventory of a working computer, the metabolism, and the ENIAC ledger of scaling levers.
Headline results, all computed in this repository
- ⁹³ᵐMo releases β = 2.87 photons per triggered decay and ¹⁷⁸ᵐ²Hf releases β = 12.39, measured from ENSDF intensities rather than folklore.
- ²⁴²ᵐAm satisfies every keystone inequality at ordinary reactor flux with library data: the keystone gate exists at reactor scale, and the research program is therefore miniaturization, not existence.
- The digital twin samples its exact Boltzmann law at 26 decays per independent sample, 945 times cheaper per sample than the magnetic tunnel junction benchmark at the 8.4 eV carrier.
- A divide by N counter turns the Poisson stream into a clock with jitter exactly 1/√N; a written bit in a lossy loop dies below loop gain 1 and latches above it, which is the keystone criterion measured as the memory regeneration requirement.
- A valve keyed emission line carries reliable data through a sealed wall with no antenna, no cable, and no penetration.
Reproduce everything
pip install -r requirements.txt
python figures/make_figures.py && python gates/isomer_screen.py
python transport/gate_demos.py && python transport/degree_check.py
python gates/edm_plan.py && python simulator/nuclear_ising.py
python simulator/components.py && python transistor/make_transistor_figure.py
python transistor/make_datasheet_figure.py && python transistor/sealed_unit.py
Every figure, catalog, and results file regenerates byte identically; the reproduce workflow runs the same steps on every push.
Citation and license
Cite via CITATION.cff; this release is archived on Zenodo with a DOI minted automatically from .zenodo.json. Prose and figures are CC BY 4.0; code is MIT.
Max Freedom Pollard, born February 17, 1993 · ORCID 0009-0007-0059-3319
The supporting physics is settled. The keystone gate is proven at reactor scale and unproven at crystal scale. Shrinking it is the whole game.