Skip to content
@Zer0pa

Zer0pa

Zero-Point Architecture For Intelligent Machines

Zer0pa

Encoding, computation, and design at the frontier of information theory, computational physics, and engineered minimalism. Published in public repositories, from the south end of the map.

What Zer0pa Is

Zer0pa is a research and engineering laboratory organised as a portfolio of independent products. Each product is its own invention, its own codec or artefact, its own commercial story. The portfolio is held together by a shared philosophy, not a shared implementation.

The Philosophy — Why "Zero Point"

In physics, the zero point is the ground state — what is always there, what cannot be removed. Zero point encoding applies the same posture to information: find the minimal, fundamental representation for a signal or structure in its native domain, and ship from that frontier. The ground state of a domain is where foundational work actually happens. That is where we work.

Portfolio

The portfolio spans three active categories and one emerging.

ZPE — Encoding Products

Per-modality geometric encoding. Each product addresses a specific signal or structural domain under bounded scope, with honest negatives and real positives. Shipped products include touch, mental imagery, symbolic music, sparse-stroke image, and structural diagram encoding. Research-in-progress includes smell and several others — XR, FT, Geo, Video, IoT, Robotics, Ink, Mocap, Neuro, Prosody, Bio, Cipher. Products graduate from research to shipped when they earn it.

Taste is published as an evidenced negative — a formally tested modality where the current lane object does not sustain honest geometric structure. Negatives are first-class results.

DM3 — Deterministic Computation

A deterministic mathematical artefact whose output is invariant across the tested substrate axes of a commodity mobile SoC, attested by a NIST-KAT trust chain. Gate surface identified by canonical SHA; dynamics-layer basin distribution holds across four independent substrate conditions. The category this opens is determinism-as-a-service for mobile computation — a property of the device, not a privilege of the datacenter.

Zero Class — Design Physics

Design-physics engineering at the ship scale. Current work is the Zero Class Vessel Hull 20098 concept — a hydrogen-electric commercial freighter — and the ZeroShip platform that holds ship-design systems to scientific standards. Public surfaces publish architecture, authority, and honest boundaries without releasing private code, CAD, or partner data.

Rosetta — Ancient Language (Emerging)

Machine learning on ancient scripts including cuneiform and Indus Valley. Currently incubated within ZPE-Cipher; will graduate to its own portfolio as the work matures. Scripts that fell silent for millennia learning, slowly, to travel again.

How We Publish

Always-in-beta is our commercial posture, not a hedge. Products ship when they have utility. They improve continuously because scientific and engineering breakthroughs do not arrive on a schedule. What you install today works. What you install next month may be materially better.

Honest scope. Every product names what it does, what it doesn't, and where the boundary sits. We claim novelty where we have it. We call standard techniques standard.

Per-product novelty. One portfolio-wide license (Zer0pa Source-Available License) with per-product novelty schedules. Protection is honest — enumerated where the invention is, silent where it isn't.

Negatives are results. When a modality or approach fails under rigorous falsification, we publish that as a first-class artefact, not a setback. A field that stops paying for the same failed experiment twice is a field that gets to move faster.

Where Work Lives

Each active product is its own repository — self-contained, with its own README, CI, proof artefacts, and Verdict. The full live index follows; see each repo's README for headline metrics, comp benchmarks where applicable, and current state.

ZPE — Encoding Products (17 public lanes)

  • ZPE-Bio — Deterministic ECG codec; Python + Rust core; 100% integrity across MIT-BIH, NSTDB, EDB, PTB-XL
  • ZPE-Diagram — Deterministic structural-with-style diagram codec; SVG line-graph transport
  • ZPE-FT — Deterministic financial time-series codec; pattern search; parquet+zstd+DuckDB comparator
  • ZPE-Geo — Deterministic geospatial trajectory codec; H3 indexing; Douglas-Peucker comparator
  • ZPE-Image — Deterministic sparse-stroke image codec; geometry-layer encoding for glyphs and skeletons
  • ZPE-Ink — Deterministic digital-ink codec (.zpink); CRC-framed lossless stroke roundtrip
  • ZPE-IoT — Deterministic IoT sensor codec; 1D time-series; Rust core + PyO3 bindings; zstd/lz4/zlib/gorilla comparators
  • ZPE-Mental — Deterministic endogenous visual form codec; Klüver taxonomy (tunnel, spiral, lattice, cobweb)
  • ZPE-Mocap — Deterministic motion capture codec; CMU corpus; ACL comparator
  • ZPE-Music — Deterministic symbolic-score codec; MusicXML roundtrip with note-local expression refinement
  • ZPE-Neuro — Deterministic neural signal codec; extracellular spike events; DANDI-anchored
  • ZPE-Prosody — Deterministic speech prosody codec; F0 / energy / duration / voiced-mask contour bundles
  • ZPE-Robotics — Deterministic motion kernel; Wire-V1 transport; replay + search-without-decode; LeRobot benchmark
  • ZPE-Smell — Deterministic smell codec; surrogate receptor-response encoding (research-in-progress)
  • ZPE-Taste — Bounded-negative evidence record (the lane object did not sustain honest geometric structure on the committed evaluation)
  • ZPE-Touch — Deterministic touch stream codec; contact + thermal + vibrotactile + proprioceptive branch isolation
  • ZPE-XR — Deterministic XR hand-stream codec; pose compression; ContactPose benchmark

Cipher and Video are in private development; Rosetta is currently incubated within Cipher and will graduate when the work matures.

DM3 — Deterministic Computation

  • DM3 — Deterministic mathematical artefact; substrate-invariance across mobile SoC tested axes; NIST-KAT trust chain

Zero Class — Design Physics

Infrastructure

  • .github — Organization profile and community health files
  • workflows — Reusable GitHub Actions workflows (release pipelines, validation surfaces)

Zer0pa is a private South African deep-tech company operating as a laboratory. We publish because verifiable science in public repositories is the funding model we believe in — and because the instruments required to do foundational work no longer respect the old maps.

architects@zer0pa.ai · zer0pa.ai

Pinned Loading

  1. ZPE-XR ZPE-XR Public

    ZPE-XR V0.0: DETERMINISTIC XR HAND-STREAM CODEC: Hand Tracking | Pose Compression | ContactPose Benchmark | Gesture Transport | Rust-Backed Wheel

    Python

  2. ZPE-Robotics ZPE-Robotics Public

    ZPE-Robotics V0.0: DETERMINISTIC MOTION KERNEL: Wire-V1 Transport | Replay | Search Without Decode | Anomaly Detection | 187x Real-Data

    Python

  3. ZPE-Neuro ZPE-Neuro Public

    ZPE-Neuro V0.0: DETERMINISTIC NEURAL SIGNAL CODEC: Extracellular Spikes | Neuropixels | DANDI-Anchored | IBL Validated | Spike Train Transport

    Python

  4. ZPE-Geo ZPE-Geo Public

    ZPE-Geo V0.0: DETERMINISTIC GEOSPATIAL CODEC: Trajectory Compression | H3 Hexagonal Roundtrip | Maneuver Search | Fidelity Validation | AIS Baseline

    Python

  5. ZPE-IoT ZPE-IoT Public

    ZPE-IoT V0.0: DETERMINISTIC IOT SENSOR CODEC: 1D Time-Series | Chemosense | 17x Mean Compression | Rust Core | PyO3 Native | Edge-Deployable

    Python

Repositories

Showing 10 of 23 repositories

People

This organization has no public members. You must be a member to see who’s a part of this organization.

Top languages

Loading…

Most used topics

Loading…