Question 1

What does the Check Your Stack assessment measure?

Accepted Answer

Twelve runtime governance invariants — relationship-aware access, corpus ontology, scoped retrieval, prompt compilation, reasoning and uncertainty, per-claim grounding, deterministic gating, signed Claim Ledger, signed audit replay, ARBV boundary verification, waiver governance, and cryptographic execution envelope. For each, the question is falsifiable: can your stack produce the artifact the invariant promises? "Not sure" counts as "cannot produce" in audit and procurement.

Question 2

How does the assessment use my answers?

Accepted Answer

Each answer is weighted (yes = 1.0, partial = 0.5, unsure = 0.2, no = 0.0) into a 0–100 score that maps to a buyer-facing tier: Packaging boundary only, Observable RAG, Partial evidence boundary, or Runtime governance. Missing and partial proofs are clustered by the Kenshiki layer that closes them. The result is a proof plan ranked against the regulatory horizon — what to fix first and why.

Question 3

Why does "not sure" count as missing?

Accepted Answer

Audit, procurement, and discovery treat absence-of-records as evidence of unreadiness. A team that cannot say with confidence whether the stack produces an artifact will not be able to produce it on demand. Conservative weighting reflects the procurement reality, not pessimism — the question lands when someone outside asks for the record, and "we'd have to check" has the same operational consequence as "we don't have it."

Question 4

Is the assessment public or shareable?

Accepted Answer

Yes. Answers persist in the URL so the resulting tier, plan, and regulatory pressure ranking can be shared with internal teams, auditors, or counsel without re-entering responses. The assessment runs entirely in the browser; no answers leave your machine.

Question 5

What do I do with the result?

Accepted Answer

Each missing proof is mapped to (1) the regulators that demand it, (2) which existing tools help, and (3) the Kenshiki layer that closes the gap. The proof plan is ordered by regulatory urgency, so the highest-pressure gaps surface first. Use the result as input to architecture, vendor diligence, or board-level AI risk reviews.

Can your AI stack defend one decision?