MOAT

Built to be irreplaceable, but only where the code actually proves it.

This surface audits the six moat claims against the real VetIOS codebase. It keeps the strong parts, marks the partial parts honestly, and calls out the missing layers that still need product and infrastructure work.

OverviewModel CardsDevelopersPassive SignalsNetwork LearningPetPassEdge Box
Public model cardsDeveloper portalOpen console
Implemented
1
Partial
5
Missing
0
Platform Coverage

What the product already covers

Consumer layer
PetPass (pet owner app)
Partial

PetPass now has real owner accounts, pet links, timeline entries, and notification deliveries behind the preview surface.

Open surface
Clinic PetPass integration
Partial

Clinic-owner links, consent capture, and owner-pet relationships are now wired as network infrastructure.

Health history & alerts
Partial

Owner-facing timeline and alerts now have database-backed timeline and notification records, though full mobile distribution still needs productization.

Open surface
Clinical layer
Inference Console
Implemented

Structured, free-text, and JSON inference workflows are live.

Open surface
Outcome Learning
Implemented

Clinician validation, outcomes, and learning feedback loops are live.

Open surface
Adversarial Sim
Implemented

Simulation and adversarial stress workflows are already exposed in-product.

Open surface
Data & ML layer
Clinical Dataset
Implemented

Canonical case storage, curation, and export flows are present.

Open surface
Experiment Track
Implemented

Dataset versions, hyperparameter records, model lineage, and comparisons now operate as one reproducible AI research stack.

Open surface
Model Registry
Implemented

Governed registry, promotion controls, rollback, and routing already exist.

Open surface
Infrastructure
Telemetry
Implemented

Execution telemetry, observability, and live dashboard signals are wired.

Open surface
Network Intelligence
Implemented

Topology, routing, and governance-linked network intelligence are present.

Open surface
Edge Box (offline)
Partial

Edge boxes now have a registry, sync jobs, and staged artifacts, though full field deployment and device auth still need hardening.

Open surface
Moat Audit

Six structural moat layers

Each claim below is tagged by what exists today: implemented, partial, or missing. The goal is to make the moat operational, not just presentational.

NVIDIA

VetIOS Developer API

CUDA ecosystem lock-in

Partial

PIMS vendors build on it.

Available now
Authenticated API routes already exist for inference, outcomes, simulation, evaluation, and passive connector ingest.
A public developer portal, published API products, and endpoint catalog now document the integration surface.
Self-serve onboarding intake and admin-issued partner machine credentials now exist in the control plane.
Still missing
Partner billing, quotas, and stronger lifecycle analytics are still not implemented.
Public docs are still catalog-first rather than a fully versioned external developer contract.
Developer PortalDeveloper OpsInference Console
APPLE

PetPass app

Two-sided consumer ecosystem

Partial

Pet owners pull clinics toward VetIOS.

Available now
A PetPass preview app now exists with owner alerts, history timeline, and clinic-linked care actions.
Owner accounts, pet links, clinic-owner links, consent records, timeline entries, and notification deliveries now exist as real infrastructure.
Still missing
Invitation flows, consumer authentication, and mobile-grade owner distribution are still not implemented.
The network substrate is real now, but the end-user app experience is still early rather than mass-market ready.
PetPass Preview
AWS

Federated Outcome Learning

Data flywheel -> compounding moat

Partial

The network gets smarter every case.

Available now
Inference -> outcome -> dataset -> benchmark -> promotion loops are implemented.
Experiment Track now stores dataset versions, hyperparameters, model lineage, and comparison evidence so published results can be independently reproduced.
Outcome-linked learning cycles, calibration, adversarial evaluation, registry promotion, and a public learning snapshot are live.
Federation memberships, tenant snapshot publishing, and weighted aggregation rounds now exist as real infrastructure.
Still missing
The current federation layer is coordinator-driven rather than privacy-preserving secure aggregation.
Automated clinic enrollment, round scheduling, and hardened governance over cross-clinic participation still need productization.
Network LearningExperiment Track
OPENAI

Inference API

Middleware everyone routes through

Implemented

All vet AI agents call VetIOS.

Available now
Inference routing, telemetry, clinical integrity checks, and registry-aware model selection are live.
The core POST /api/inference surface is already production-shaped.
Still missing
Broader agent ecosystem adoption still depends on external integrations, docs, and partner distribution.
Inference Console
TESLA

Passive Signal Engine

Passive fleet data collection

Partial

Clinics generate data by working.

Available now
Passive connector normalization exists for lab results, rechecks, referrals, imaging, and medication refill signals.
Episode reconciliation, a clinic workflow signal dock, and a published connector catalog are already built.
Still missing
Most connector ingestion is still manual or shared-secret based, not turnkey vendor sync at fleet scale.
There is no connector marketplace, scheduler, or broad EHR/PIMS sync coverage yet.
Passive Signals
ANTHROPIC

Public Model Cards

Transparency as trust moat

Partial

Auditable, certified, trustworthy.

Available now
Registry governance, gate status, blockers, and lineage already exist internally.
A public read-only model card surface is added in this change so the registry can be shared externally.
Publication records, certifications, and attestations now exist behind the trust surface.
Still missing
Third-party signed attestations and automated evidence ingestion are still not implemented.
Public Model CardsTrust OpsRegistry Control Plane
Consumer pull
PetPass now has real owner-network infrastructure behind it, but the full consumer distribution and app adoption loop still need productization.
True federation
Learning loops exist and the new public network-learning surface exposes them, but cross-clinic privacy-preserving federation is still future work.
Partner platforming
Self-serve onboarding and partner keys are now built, but quotas, billing, and richer partner lifecycle analytics still need productization.
Trust publishing
Publication, certification, and attestation records now exist, but signed external evidence ingestion still needs hardening.