Get Started · L0 is live · 2026-06
Start here. Honestly.
The first public release is live. The genmount client is on PyPI (pip install genmount) and three open-weight education models — TCM, Ayurveda, and Tibetan medicine — are on Hugging Face (gated, free registration). Everything runs locally on your own machine via Ollama; no account is needed for local use. The Docker image, desktop app, and hosted trial demo are still on the way — and there is no hosted SaaS with real medical data, ever. This page tells you what is real today and what is still on the horizon.
We refuse to ship a product before its Spec is honest.
✅ L0 live · pip install genmount · Spec available now§1 · Where we are
What is real today vs what is on the horizon
Two clear columns. We do not blur them in pitch decks and we do not blur them here.
Available now (2026-06)
- genmount client on PyPI · pip install genmount · runs locally, no account needed
- Three open-weight GGUF education models on Hugging Face · TCM / Ayurveda / Tibetan medicine · gated, free registration
- Ollama local path · download a model and run it entirely on your own machine
- Spec Portal · five core pages (Cosmology / Pillars / Propositions / Gates / Robotics)
- Paper v0.2 on Zenodo (DOI 10.5281/zenodo.20053545, 2026-05-06)
- Industry-position disclosure · claim-tier paradigm challenger, not foundation-lab tier
- Voice Baseline G1–G15 published · current status 0 / 15 passed
On deck · not yet shipped
- Cloud fallback + model sync · authenticated routing for hard cases (rolling out)
- Docker image · not yet built or pushed
- Desktop app · macOS / Windows / Linux — design only
- Hosted trial demo (the L0 tier below) · feature-limited, 24-hour data deletion — not yet live
Honest-status disclosure tracks Cosmology · §4 (truthfulness clause) and Gates · status snapshot. We will not move an item from the right column to the left without engineering Gate 1 hardening + a §11 paper gate passing.
§2 · Four-tier distribution
Where your data lives, when it ships
Four tiers are pinned. The Get Started CTA defaults to L1 — your local install. We never default to a hosted tenant, and we never run multi-tenant hosted with real medical data.
| Tier | Audience | Form | Data lives in | Stage |
|---|---|---|---|---|
| L0 | Trial demo Curious users | Hosted web demo, feature-limited | Our infra · explicit disclaimer · 24-hour data deletion | Not yet live |
| L1 | Personal TCM enthusiasts / individual developers | Ollama / desktop app / Docker | Your local machine | Live now |
| L2 | Institution TCM clinics / rehab centres | Docker on customer's own server | Customer's own infrastructure | v1 pilot |
| L3 | Enterprise Hospitals / groups | We deploy on-site in your machine room | Customer's machine room | v2+ |
Permanent red line
We will never run a multi-tenant hosted product with real medical data. This would violate crypto-shred + the two-project hard boundaries + medical compliance, and it would dissolve the anti-scale paradigm-challenger position we hold.
Term discipline · sub-domains + software isolation = multi-tenant SaaS. That is not a private cloud. A real private cloud requires dedicated hardware + customer-held keys + portability terms + quarterly third-party audit — all four or we do not use the term.
§3 · Three local paths
Three ways to run it locally
All three keep your data on your own machine. The Ollama path is live today; the desktop app and Docker image are still on the way.
Ollama integration
Livepip install genmount + pull a model
Open-weight GGUF models on Hugging Face plus an Ollama Modelfile, wired up by the genmount client. Runs next to whatever local LLM you already have; nothing leaves your machine.
Desktop app
Not yet liveOne-click install, no terminal
macOS / Windows / Linux desktop bundle. Local model, local audit log, local replay. Designed for TCM practitioners and rehab clinicians who do not run a terminal.
Docker image
Not yet livedocker compose up — single host
Single-host Docker image plus compose file. Same engine as the desktop bundle, intended for developers and institutions running their own server.
§4 · The best way in today
Read the Spec — about thirty minutes
Five pages cover the worldview, the engineering pillars, the four nails, the gate map, and the embodied architecture. Read in this order:
- 01
Cosmology
Why this is an interpretive world model rather than a bigger sequence model. Five-stage reasoning chain, four admission gates, honest-status disclosure.
- 02
Pillars
Four engineering pillars — Xirang (HSS state topography), Chonglou (tri-tier permanence), Zhiju (bicameral decision engine), Lianshan (multi-hop discovery).
- 03
Propositions
Four nails that must all appear together in any pitch: bounded output, health closed loop, refusal brain, research moat.
- 04
Gates
Eleven-gate governance map: three paper §11 gates + four engineering gates + four admission gates, with the current status snapshot.
- 05
Robotics
Five-layer embodied architecture with the bicameral capability + boundary tracks. Nine hard boundaries. The only two battlefields we touch: TCM-assist and rehabilitation-assist.
We do not ship docs ahead of code. We do not ship code ahead of its Spec. The Spec is what you can verify today.
Local client + open-weight models: live today. Docker, desktop, hosted demo: on the way. Hosted SaaS with real medical data: never.