Gates Map · 11-Gate governance overview
11 gates = 3 + 4 + 4
The 3 paper gates are the foundation (underground). The 4 engineering gates are the structure (above ground). The 4 admission gates are the entrance (the front door). Three sets, different roles, well-defined dependencies — and they cannot be interchanged.
Foundation un-poured: you may still push the engineering gates, but you may not go live.
§1 · The full picture
Three semantic layers · one diagram
The 11 gates sit in three different semantic layers — not sub-categories of one set. Bypass any layer and the whole claim is disqualified.
Set 01 · 3 gates
Paper §11 · three gates
Foundation · falsifiability
❌ all three not yet passed
PAPER v0.2 §11
Set 02 · 4 gates
Engineering · four gates
Engineering threshold · in progress
🟢 Gate 1 (all 23 items) skeleton complete
INTEGRATION-LAYERS §4
Set 03 · 4 gates
Admission · four gates
Entrance · onboarding external medical systems
⏸ on standby / see Cosmology
WHY-COSMOLOGY §3.1
§3.1 · Paper §11 · the foundation
The three gates are the falsifiability foundation — non-bypassable
If any one gate fails, the technical claims of the paper remain falsifiable. All three currently fail.
§11.1
Interpretability
Any conclusion must trace back along the five-stage chain to a primary-form anchor.
Pass criterion
Engineering Gate 1.B fully passes + third-party expert blind review ≥ 80% agreement
§11.2
Reproducibility
Same input, same environment → same output.
Pass criterion
30 classical cases reproduced at 100% + 5 independent labs replicate
§11.3
Falsifiability
There exists a clear set of conditions under which the framework can be refuted.
Pass criterion
Publish ≥ 10 "if X is observed, the framework fails" statements
Current state
❌ all three not yet passed (v0.2)
§3.2 · Engineering gates · in progress
Engineering thresholds · the main line
Strict dependencies between gates. If any prerequisite fails, the subsequent gate is not allowed to start.
API end-to-end stability
🟢 all 23 items skeleton complete
MCP read-only entry
⏸ awaits full pass of Gate 1
MCP extension
⏸ awaits Gate 2 + any §11 gate
Governance alignment
⏸ awaits Gate 2
§3.3 · Admission gates · the entrance line
Onboarding external medical systems
Before adapting any new system — veterinary, phytomedicine, Ayurveda, others — four gates must pass. Detailed criteria on the Cosmology page · §3.1 Admission tests.
- G1 Root-anchor mapping — ≥ 20 core syndromes, coverage ≥ 80%
- G2 Baseline consistency — projectable onto the six-qi baseline
- G3 Reversible differentiation — ≥ 5 cases, 100% reversible along the five-stage chain
- G4 Boundary declaration — projection surfaces (human / animal / plant / mind …) explicitly declared
§4 · Current status snapshot
Gate-pass state (2026-05-09)
Any gate-status change must be reflected in this table and in the related sub-document. No embellishment.
| Gate | Status | Evidence / notes |
|---|---|---|
| §11.1 Interpretability | ❌ Not yet passed | — |
| §11.2 Reproducibility | ❌ Not yet passed | — |
| §11.3 Falsifiability | ❌ Not yet passed | — |
| Engineering Gate 1 (1.A–1.E) | ✓ Passed / skeleton complete | All 23 items skeleton complete (2026-05-09). Full delivery still awaits external dependencies (≥ 20 classical cases + ≥ 10 clinical records + ECS hot-path tests + clinical LOI). |
| Engineering Gate 2 | ⏸ Awaiting prerequisite | Awaits any §11 gate + Gate 1 engineering hardening |
| Engineering Gate 3 | ⏸ Awaiting prerequisite | Awaits Gate 2 + any §11 gate |
| Engineering Gate 4 | ⏸ Awaiting prerequisite | Awaits Gate 2 |
| Admission Gates 1–4 | ⏸ Awaiting prerequisite | Awaits an external medical system to apply |
§5 · Industry positioning
Declared Paradigm Challenger
Narrative complete, but key thresholds of the 11 gates remain uncrossed — so the "challenger" identity is, today, a falsifiable proposal, not a validated paradigm.
Three tiers exist simultaneously and the gaps between them are wide. Only the composite tells the truth.
Layer 01
Engineering / compute
Tail end
We do not train foundation models — only adapters on top of Llama 3.2. No GPU cluster, no frontier infrastructure. We are not running this race. That is not a defect; it is the precondition of the anti-thesis.
Layer 02
Brand / community
Early release
doorm-ai org with three public repos; paper v0.2 on Zenodo (DOI 10.5281/zenodo.20053545); users / citations / deployments currently all at 0.
Layer 03
Paradigm / thought
X-axis challenger
We chose the anti-scale side and stand apart from the Hinton / Feifei / Sutskever mainline. Comparable on the same axis: LeCun JEPA, Friston Active Inference, Hawkins Thousand Brains, Marcus neuro-symbolic — same character, very different mass.
Composite verdict
"Declared" challenger ≠ "Established" challenger
Held (4 ✅)
- Complete paradigm narrative
- Clear X-axis (anti-scale)
- Firm anti-scale stance
- Third-party public release (Zenodo DOI live)
Blocking "established" (6 ❌)
- Paper §11 — three gates not passed
- VOICE admission 0/15
- CDSS production rubric (7 items) — postponed
- No clinical / pilot evidence
- Users / citations / deployments all at 0
- Capital / compute / academic standing — orders of magnitude behind LeCun / Friston / Hawkins
Upgrade path · 6 objective thresholds. At least 3 must hold concurrently before this verdict can be re-written.
- C1 Paper §11 — all three gates passed
- C2 VOICE admission ≥ 12/15
- C3 Independent third-party reproduction of the core demo (consumer GPU)
- C4 ≥ 1 small-scale TCM / rehab pilot (IRB approval + real usage logs)
- C5 ≥ 1 anti-scale-camp scholar publicly cites or replies
- C6 ≥ 1 independent engineering team onboards a new medical system through the governance admission gates
§6 · Voice Interaction Baseline
Voice channel · 🔴 not yet on-ramp · 0/15
All 15 admission criteria must pass → triggers the formal unfreeze of CDSS Layer 2. Any failure → "voice available" must not be claimed publicly.
The voice channel today is a skeleton — meta-layer rules, type contracts, and negative guards are in place, but the acoustic engine, the dialogue engine, and the warmth generator are not. We can guarantee "do no harm" (no fake emotion detection, no mysticism), but we cannot yet guarantee "feels like talking to a person".
Admission overview
🔴 0/15
Goal
🟢 15/15 → triggers CDSS Layer 2 formal unfreeze
Convenience
🔴 Not present- VOICE-G1/voice/ingest endpoint connected to a real ASR (model path through Finetune Llama 3.2 adapters)
- VOICE-G2extract_prosody_from_audio no longer returns confidence=0.0 placeholder; real acoustic feature extraction in place
- VOICE-G3Any supported client (web / mobile / desktop / watch) can start a voice dialogue in ≤ 3 steps
Latency
⚠️ Unmeasurable- VOICE-G4End-to-end P50 ≤ 1.5 s and P95 ≤ 3.0 s, stable on a standard benchmark set
- VOICE-G5Streaming responses enabled — the "user finishes full sentence → waits → full reply" mode is not allowed
Accuracy
⚠️ Partly right, partly coarse- VOICE-G6ASR character error rate ≤ public baseline in Mandarin + at least one other published language
- VOICE-G7Tonic / wuyin labels reach human-annotation kappa ≥ 0.6 on ≥ 200 real (not synthesised) voice clips
- VOICE-G8valence_hint uses a backend-trained model, kappa ≥ 0.5 against human annotation
- VOICE-G9cross_paradigm_anchor traces every wuyin back to a primary-form anchor — if the trace fails, the input is rejected
Flow
🔴 Not present- VOICE-G10A stage-aware dialog manager exists: one interaction traverses the four explicit phases — synthesise / analyse / discover / resolve — each with its own record_span
- VOICE-G11Downstream prose modulator consumes prosody_features with auditable differences
Warmth
🔴 Only negative guards- VOICE-G12Persona consistency test passes — the same speaker recognisable as the same dialogue subject across ≥ 10 different-topic conversations (address style, register, pacing)
- VOICE-G13Empathy phrase template library covers ≥ 6 everyday situations (greeting, apology, care, encouragement, closing, waiting); a valence-aware tone modulator is wired in with auditable mapping
- VOICE-G14Contextual memory participates in responses — within one session the system can refer back to earlier mentions; memory routes through core/memory_schemas.py three-tier memory
- VOICE-G15Reverse user study with ≥ 20 people in ≥ 5-minute conversations: median "feels like talking to a person" score ≥ 4/5 (pre-registered survey, no cherry-picking)
The 11 gates are self-verification. The 7 production-grade CDSS items and the three-quadrant Entropy Map are external / meta-level verification. True delivery = all three sets satisfied — bypass any one and the claim is disqualified.
We walk on the foundation, not on the spire.