Capability readiness

How this wiki labels what is actually true about a capability, so that a reader can tell a public repository from an unverified prototype at a glance. This is the wiki’s main structural defence against a plausible-sounding but misleading capability list.

The problem

Capability maturity varies a lot — public repository versus “research prototype, code not verified.” Do not present a capability table as a single validated product suite; keep the per-row evidence pointer visible.

That instruction from the source material is a safeguard, and it governs Capability portfolio directly.

The readiness labels

LabelMeaningExample
Public codeAn identifiable public repository existsExordium, LinMulT, BlinkLinMulT
PublishedA peer-reviewed publication or formal record supports the claimThe ADOS-2 study, the AIRS paper
Research prototypeDemonstrated in a research context; maturity not verifiedSkel3D (repository exists, models/instructions incomplete)
HistoricalDocumented past work, not claimed to be maintained or deployable todayLabelMovie, the AAC recommender
OpportunityDiscussed and plausible; no agreed plan, no commitmentCamera-based therapy monitoring, the Danish data route

The five dimensions

Readiness in one dimension implies nothing about the others:

DimensionQuestion
Clinical validityDoes the measurement track what it claims to track?
Clinical utilityDoes using it change decisions or outcomes?
Intellectual propertyWho owns what, and on what terms?
Regulatory statusWhat is it permitted to be used for?
Commercial readinessCould it be sold, supported and adopted?

How this shows up in practice

  • Speech Gap has strong IP and demonstrated sensitivity, and an untested prognostic application. Those coexist without contradiction.
  • Argus Cognitive has an FDA-cleared product in cardiology and a separate autism product. Clearance for one is not clearance for the other.
  • The Danish route is a genuine opportunity with no data access established.

Capability portfolio · Clinical AI translation · Editorial safeguards · NIPG