Current review: 9 September 2026. Eve contains 25 guides and 217 slides. The library contains 87 guides, 1,247 unique slides and 1,270 guide-slide instances across 605 cited source paths. Earlier delivery counts in README and TODOS are historical checkpoints.
| Concern or requested enhancement | Delivered result | Where to inspect it |
|---|---|---|
| A category column was labeled “Question” | Comparison headings are authored for their actual fields. The label remains only where the corresponding cells contain questions. | Progress systems |
| The third row was always highlighted | Peer rows have equal weight. A highlight requires a named subject and an explanation in the slide brief. | Progress systems |
| Eve's features were too shallow | Dedicated chapters teach product graphs, coverage and impact, tasks, decisions, delivery, walkthroughs, operator investigation, Studio analysis/actions and member capabilities. | Eve's starting guide |
| Feature graphs needed depth | Eight slides cover containment, typed links, cross-links, live overlays, the implemented Explorer, bounded neighborhoods, matching rules and an exercise. | Product graph |
| Task tracking needed depth | Nine slides cover actionable records, kinds, preparation, delivery states, recovery, editing/linking, the board, priority/ownership and review. | Work tracking |
| Analysis needed depth | Coverage has eight slides; Studio analysis has ten, including concept dossiers, evidence, quality, safety, cost, media governance and readiness evaluation. | Coverage analysis, Studio analysis |
| Repeated tables and generic columns | The Eve design pass added concrete cases, typed relationships, source-shaped records, recovery diagrams, gated decisions, an honest progress chart and real React-component captures with declared fixtures. Useful field comparisons remain tables. | Eve design review, delivery evidence |
| Sequence and organization | The route establishes product evidence before work, decisions before delivery, specialized surfaces before the complete practice, then architecture and its dated runtime companion. | Guide-by-guide sequence decisions |
| Remaining Living Scenes overview defects | Four old column layouts now use responsibility/surface comparisons and policy/promotion branches. The explanations distinguish local fixtures, helper results and a venue publication request. | Runtime guide, fix package |
| Technique and primitive depth | Twenty-five slides teach all twelve catalog records, all nine primitive parameter contracts, three distinct continuity layers, complete accessibility substitutions and the actual shallow-freeze/version boundary. | Technique catalog, source map |
| Compatibility and admission depth | Nineteen slides teach four policy stops, seven equal-weight dimensions, exact arithmetic and thresholds, all sixty template cells, tone separation, filter behavior and the bounded override helper. | Compatibility guide, delivery receipt |
| Consumer, render and distribution depth | Forty-seven slides trace the actual card and compose surfaces into keep/share authority, tenant policy, render admission, durable media, public resolution, provenance, accessibility and release gaps. Five real UI captures distinguish implemented, unwired and denied states. | Consumer guide, delivery receipt |
| V1 shared-platform depth | Two linked guides add sixty-three slides for contracts, partial results, authorization, idempotency, tenant propagation, residency, persistence, erasure, transport selection, Redis event delivery, durable queues, live search, recommendations and recovery. | Contracts and tenancy, events, jobs and discovery |
Living Scenes composition now has two separate learning sequences: the member composition workflow and Compose Assist contracts and evaluation. Thirty-one new slides include five implemented-component crops, two explicit adjacency diagrams, six decision diagrams, source-shaped records and a worked five-case evaluation. The guides distinguish a local rehearsal from actual model dispatch, durable persistence and media rendering. Their delivery receipt records the scoped source, browser, PDF and narration evidence. All previously published slides and recordings, including Eve, are preserved.
The technique and compatibility guides extend that route between composition and Compose Assist. The consumer guide follows Compose Assist and governance so its surface audit can name the mechanism and policy each caller depends on. Ninety-one slides use complete contract tables for closed sets, relationship diagrams for dependencies, gated flows for policy branches and worked cases for scorecard and compatibility arithmetic, plus five actual UI captures. Fifty-one complete handbook units now have reviewed source assignments across the two packages. The slides state that catalog freezing is shallow, tone membership is a separate helper, enforcement strings do not prove timing or callers, and renderer, persistence and media acceptance sit outside the technique validator. The consumer guide then names the implemented downstream boundaries and the callers and acceptance evidence that are still missing.
The current Eve artifact audit checks every slide's generated title and reading edition, removes artificial numbering from peer columns, and verifies comparison headings and authored emphasis. The original 204-slide design pass has retained browser/PDF evidence; thirteen later operations additions have their own delivery receipts. This is not a new human comprehension study or a claim that every feature is current in a deployed Eve service.
The V1 platform-foundations route now starts with the existing architecture overview, continues through the contracts and tenancy detail, uses the existing communication overview as a bridge, then enters the events, jobs and discovery detail before Sophia, Iris and Psyche/Lilith. Fifty-six meaningful source units are fingerprint-bound to the exact slides that teach them. The route explicitly distinguishes source behavior from a mounted route, deployed dependency and accepted user journey. It also corrects the live recommendation composite to the implementation's five actual signals and fixed weights.
The whole-library design overhaul is unfinished. Section 10 of TODOS still tracks the complete benchmark collection, typography and optical review, quantitative provenance, cross-slide continuity, representative output imagery, actual-size readability, reader comprehension and editorial subtraction across every guide. Extensive UI/UX walkthroughs, remaining V1 source coverage and final whole-library narration review also remain open. Human reader sessions and normal-speed listening must not be reported as performed when they have not taken place. The V1 workstream records the remaining scope.