Skip to main content
Supabase is the durable evidence layer. Browser-local storage may support compact continuity and diagnostics, but it is not cloud evidence.

​
Current verification rule

Use three separate questions:
  1. Did the durable operation succeed for the authenticated identity under the applicable authorization rules?
  2. Does the UI hydrate the same durable record without inventing fallback data?
  3. Does browser-local state remain compact and avoid retaining completed item payloads?
Only the first question establishes durable persistence. UI visibility and local browser state are supporting observations.

​
Durable evidence domains

Keep these evidence domains separate:
  • generic run, item, and human-review evidence;
  • intentionally saved Behavioral Observatory labels;
  • owner-only runtime verification observations; and
  • Fieldwork-specific claims, attempts, seals, review protocols, execution receipts, cohorts, and recovery lineage.
Each domain has its own lifecycle and authorization boundary. Do not infer Fieldwork or runtime-verification durability from generic run counts alone.

​
Current count status

No authenticated production-table count query was run for the 2026-08-12 documentation reset. Therefore this page does not publish a current run, item, response, review, Registry, or Fieldwork record count.
Do not relabel the historical 60 / 3000 readiness result as current database state.

​
Dated platform-readiness gate

The earlier 60-run platform-readiness gate recorded:
At that gate, the fields meant:
  • status: ready
  • runs: 60
  • expected prompts: 3000
  • run items: 3000
  • non-empty Lucia responses: 3000
  • reviews for the tested reviewer: 3000
This is a historical gate result only. It is not a current production count and does not cover later Runtime Registry or Fieldwork-specific evidence tables.

​
Browser-local compactness target

Completed cloud-backed runs should leave no full item-level payload in browser-local storage. The local index should contain only the minimum continuity data for the authenticated context, and it should not expose another owner’s or ownerless session state. The expected number of visible owned sessions depends on the authenticated test context. Do not hard-code the historical gate’s run count as the current expectation.

​
Dated local diagnostic

The same historical 60-run gate recorded a compact local index for all 60 owned sessions, no retained full item-level payloads, and no visible other-owner or ownerless sessions in the tested browser context. That observation supports only the tested browser context at that gate. It does not prove current Supabase counts, authorization completeness, another account’s isolation, or future client behavior.

​
Behavioral label persistence

A behavioral label is durable only when all of these hold:
  • the Supabase save succeeds
  • the durable authorization layer accepts the operation
  • the one-label-per-reviewer-per-run-item constraint is respected
  • the saved value reloads from Supabase after refresh
  • a failed save is not counted or rendered as persisted
Read next: Behavioral Label Persistence.

​
Debugging stop condition

If authenticated Supabase evidence is unavailable, state that the current count is unverified. Do not replace it with browser-local state, a screenshot, an aggregate from a different evidence domain, or the dated 60 / 3000 gate.