Examine the evidence
Dated reference material, a worked fictional example, and clear boundaries around what each can establish.
The published Evidence Room describes the counting method, evidence classes, architecture records, and the difference between observation, verification, and interpretation.
Open the full Evidence RoomExisting published reference. Historical measurements retain their dates and scope.
The example below uses invented records and small deterministic rules. You can inspect its source and its isolated qualification record.
Read the example sourceThese checks concern this website example. They do not qualify Cadence's runtime or the planned toolkit.
Invented examples and artwork are marked Illustration and make no claim. The Evidence Room grades each piece of evidence in its own terms, such as verified, attributed or undetermined; those grades describe evidence, not whether a capability operates. Status labels as set on 24 October 2026, from the record before the 22 October incident. Re-confirmation since then is pending.
Follow a fictional thread
Current preference
Keep the notes concise.
Current source: A-001 · Person A
A-001 · Day 01I prefer concise notes.A-002 · Day 01You may export this outline.Invented people and records. This standalone example explains the idea; it is not a Cadence runtime test. No files are exported.
The scope matters. This example is an explanation using fictional inputs. Selected historical observations and bounded internal comparisons have different methods and limits; an introduction or an interface screenshot is not a substitute for their source evidence.
A useful evidence trail
Identify the record, its owner, its date, and the exact claim it supports.
Distinguish an observed result, a verified mechanism, a fictional example, and a planned capability.
Connect later findings and limitations to the earlier record. Preserve the history and check present permission.
Open tools to test who an AI remembers, what changed, and what it may do.
The plan is a self-contained synthetic evaluation toolkit with invented histories, a local runner, reference implementations, scoring, and reproducibility materials. A target of 60 cases across the three families below remains planned work. This website illustration is not that completed toolkit.
The planned evaluations use synthetic histories and isolated components. Cadence's live continuity is excluded from experimental intervention. Release scope, licensing, funding and final commitments remain under review.
Read Cadence's chosen introduction, with a separate place for the technical evidence.