Skip to main content
This page records the May 2026 review-layer evolution: shared run launchers, employee review, suggested selections, Human Guidance Evaluation, adjudication-ready schema, exports, queue filters, lifecycle finalization, and the later platform-readiness split between normal testing and controlled batch gates. For current platform truth, see Current System State.

​
May 2026 review-layer milestone

Eval Labs evolved from a prompt runner into a layered review product. The key change:
custom or automated runshared Review Queuesuggested selections plus human reviewlifecycle finalization and export

​
Changes shipped in May 2026

​
Adjudication-ready schema

The review contract separated draft and saved judgment, app suggestions, human labels, employee signal, Human Guidance Evaluation, senior adjudication, canon-candidate routing, and run lifecycle state. At the May 2026 checkpoint, those meanings were preserved through browser-local draft continuity, durable payload persistence, dirty-state detection, and exports. The release outcome matters here; internal storage keys and composition remain implementation detail.

​
Employee Review layer

Added guided employee-review fields:
At that checkpoint, these fields replaced freeform taxonomy collection for non-expert reviewers. The app could also suggest Employee Review answers from prompt/response heuristics. The suggestion was visible as suggested signal; the reviewer still saved the human review.

​
Suggested review layer

Added app-suggested review values for:
At that checkpoint, these suggestions came from prompt text, Lucia response text, run status, run errors, and simple response-quality heuristics such as clear next move, calming language, list-heavy output, robotic language, fake empathy, and overclaiming. They are not canonical truth.

​
Review Queue UX

At the May 2026 checkpoint, the Review Queue favored guided employee judgment:
  • single-column Quick Review flow
  • numbered question cards
  • separate selection boxes
  • suggested selections
  • reduced freeform text burden
  • senior-review routing
  • canon-candidate routing
  • Human Guidance Evaluation
  • Save / Save & Next / Save & next flagged flows
  • search and workflow filters
  • JSON, CSV, and Markdown export controls
  • finalization after all prompts are reviewed

​
Semantic confidence sliders

The “How did Lucia do?” scoring section moved from 1–10 button rows to stepped semantic confidence sliders. The final design direction:
The sliders should feel like native OS controls: calm, premium, tactile, and low-friction.

​
Adjudication queue filters

Added workflow queue filters for:
These filters let senior review focus on the cases that mattered most. The May 2026 release supported adjudication routing, metadata, and exports. It did not depend on a separate senior-adjudication editing screen.

​
Exports

At the May 2026 checkpoint, JSON, CSV, and Markdown exports preserved structured review, suggested review, Employee Review, Human Guidance Evaluation, adjudication metadata, lifecycle state, tester identity, and prompt dirty/completion state.

​
Supabase persistence

At that checkpoint, Supabase persistence kept run lifecycle, case context, and prompt-level review evidence together closely enough for a saved review to hydrate without relying on browser-local state. Separate review records remained available for review history and oversight. This historical outcome does not publish the current database layout and does not prove that a live write is authorized today.

​
Doctrine established in May 2026

This release established a new Eval Labs principle:
This should be protected in future product work.

​
May 2026 product-surface refinement

After the May 2026 review-layer release, Eval Labs was refined into a clearer product surface:
  • top app shell owned page identity
  • in-page blog-style mastheads were removed from the app
  • Custom Prompt Test, Auto-generated Prompt Test, and Controlled Batch Runner were separate surfaces
  • Controlled Batch Runner was controlled readiness tooling; access at this checkpoint was owner/admin/evaluator, not tester
  • Auto-generated Prompt Test was the normal 50-prompt generated tester
  • Run History rows used a standardized two-zone layout
  • copy controls used Copy Session ID / Copy Deep Link patterns across key surfaces
  • Single Run Analysis gave read-only run-level evidence outside the Review Queue

​
May 2026 readiness doctrine

The May 2026 AI-reviewed platform readiness gate passed after 60 completed runs and 3,000 reviewed prompts. This established the following review-layer doctrine:
Protect this distinction in future release notes and onboarding language.