The first chapter of the Meet Lilith track, rewritten in the Eve edition on 10
September 2026. It follows the plan in EVE-V1-REDESIGN-TODOS.md and the track
brief in meet-lilith-track.md. The receipt is under
verification/.
The job#
This is the product members actually open. In V1.0 it is Lilith on the web and installed from the web, holding four rooms — Tara, Arete, Nyx and Nisaba — and the persistent failure it teaches against is a small one: reading a message on a surface as a claim by everyone downstream of it.
Running example. One note saved in Nisaba while the connection is down. The device says saved, and the question the chapter carries is who else has said anything at all.
Three learning promises (the cover)#
- Four questions, not a funnel — route, release, capability and outcome, and why evidence for one is not evidence for another.
- One shell, rooms that own their effects — what the shared shell carries, and what it cannot promise on a room's behalf.
- What "installed" and "saved" claim — the update request, the local queue, and controls whose enforcement lives elsewhere.
Three things to remember (the close)#
- A local acknowledgement is local. The device accepted the work; nobody else has said anything yet.
- A shared facility is not a wired consumer. The shell offers a queue; the room still owes the write.
- Release scope is eligibility. It admits a room and makes no claim about how deep that room goes.
Bridge. The account panel, where a member changes a setting that four different owners have to enforce — which is chapter two.
Sections#
Three sections, opened by a section-divider that names all three.
| Section | Slides |
|---|---|
| What "it worked" has to mean | the note, the four questions, the release map, the glossary |
| The shell, and what it does not own | the shell, the two deferrals, the optional day, the update helper, the worked update |
| Where the evidence actually is | the account panel, the design rules, room maturity, the entry paths |
Treatment decisions#
- The four questions became a card grid, not a comparison table. The reviewed source rows survive exactly: each card's detail is its question and its evidence line is the third column, under the label Evidence to inspect. Outcome keeps the emphasis, because it is the only one of the four the other three can all be satisfied without.
- The deferred rooms are drawn with no edge.
lilith-v1-0became a real graph diagram, which retires nothing fromdiagrams.json— it never had an override — but it does let the release map say the thing the containment map could not: Veritas and Metis are on the slide, and no line admits them. release-enforcementbecame a compare panel rather than a third graph. Its own takeaway is a two-sided claim, and drawing three consumers of one declared cut said less than reading the two matchers across five shared questions. Re-reading the pinned middleware for this slide changed what those questions are: the gate is one request hook that runs before authentication, on purpose, so the refusal is identical for every caller; payment rails are checked first, because money fails closed before a tidy error taxonomy; and both refusals are the same 404 shape. Each matcher also has a deliberate blind spot — a room name inside a search query is untouched, and a rail with no listed prefix is not refused here at all — which the inherited "fails closed for unknown rail identifiers" phrasing had blurred.- The worked update became a decision tree with three real outcomes. The inherited gate had Apply now and Later. Later is a correct answer to the question the member was asked, and the branch a linear flow leaves out is the third one: activation was requested and the passage did not come back, which has to be offered again rather than claimed.
room-maturitylost its table. One card per room, the documented substrate as the detail and the boundary that still needs evidence on the evidence line. Nisaba carries the emphasis because the running note is saved there and its notebook lifecycle is the documented gap — the example was chosen to turn on a real one, not an invented worst case.- Three slides were kept as they were. The offline case study, the optional recurrence and the annotated service-worker excerpt are each the best available treatment of their claim, and the recurrence is the library's only one. Their prose was rewritten; their compositions were not touched.
What the rebinding needed#
Four contracts were held by slides this chapter changed, and each moved to a
slide that still uses the layout rather than being dropped: the
evidence-comparison contract to sophia-scoring-model, the relation-map
direction contract to tara-teacher-lineage, and the legacy-layout refusal in
test_composition to a legacy layout the Eve edition does not already reject
earlier. The guide's slide-brief table and composition-pilot.json both name
this guide's complete slide list, so both grew by six rows.