The Rail (V10) — Ambient Companion Architecture for V1–V9

Date: 2026-07-16

Status: Implementation active — Phase A code-complete; human exit gate open

Sources: the nine product reviews (V*/V*_PRODUCT_REVIEW_2026-07-0[67].md), V_SERIES.md, V_SERIES_PLATFORM_CONSOLIDATION.md, the 2026-07-13/14 content audits, systems/app-aphrodite.md, and a product riff on 2026-07-16. Market observations are grounded as of early 2026.

Implementation reconciliation (2026-07-22)#

This document remains the product and architecture source, but it is no longer a pre-proposal. The Phase A code path, native shell, policy kernel, channel contract, lanes, daypart programming, and pilot integrations are implemented and covered by the task ledger's automated gates. That is a repository and local-runtime statement, not a production-release claim: the week-long human dogfood exit test (HG-1), content/rights decisions, final naming, and the other human gates remain open.

Implementation resolved or clarified the following points where the original proposal was deliberately provisional:

• Product and shell ownership. D8 is now reflected throughout the code as V10. The Tauri application is the authoritative desktop shell; the shared static web/PWA build is a fallback and satellite surface, not a replacement for tray, always-on-top, docking, or native discretion behavior. V1 remains the sole identity/account authority.
• V3 venue naming. The channel is technically v3.stage, displayed as V3 Stage. Historical V3 Saraswati tenant, Unreal, and persistence names stay intact, while the unrelated @saraswati/* industrial domain remains a separate namespace. Final public branding is still HG-7.
• Shared streaming home. The audit-selected canonical substrate is libs/shared/live-media, package @oshun/live-media, rather than a new V10 stack or the proposal's illustrative libs/shared/streaming name. Aphrodite retains adult-ring policy and thin tenant adapters. Local four-tenant equivalence and supervised ingest-to-Rail playback are proven; deployed Veritas/V3 Stage hosts and the production first-party-live flips remain open, so RB.3's parent extraction item is not overstated as shipped.
• Stage media truth. Phase A uses verified progressive/VOD material and labels scheduled items as premieres. The cook, durable catalog boundary, and local UE5.5/FFmpeg release gate exist, but no fixture, plan, or declared drill is promoted into a production catalog row. Licensed third-party programming remains HG-2.
• The Thread is programming, not a channel. The morning ritual is one daypart-scoped composition in the fixed wake → Tara practice → Veritas audio → Case Files → Daily Wonder → Nyx order. Each ready item retains its owning channel authority; unavailable editions stay explicit, and audio begins only after a user press through the single audio arbiter.
• Release wording. Automated Phase A completion does not close the human or deployment gates. "Implemented," "locally verified," and "production configured" are kept distinct in the integration records and the ledger.

The task ledger at V_SERIES_AMBIENT_RAIL_TODOS_2026-07-16.md is authoritative for requirement-level completion evidence and remaining gates.

1. Thesis: the sidebar slot, and why we keep losing it#

Watch how people actually consume media during a day: work occupies most of the screen and most of the attention, while a side panel runs YouTube, a live stream, sports, music, or news. That side panel — call it the sidebar slot — is where the majority of waking media hours actually live. It is glanced at, half-listened to, and dipped into during micro-breaks. It sustains mood and provides company. It never demands that you block out everything else.

Every V-series product is currently designed as a destination: you stop what you're doing, enter, give full attention, leave. That competes for the scarcest resource a person has (contiguous free attention) and ignores the most abundant one (peripheral presence across an 8-hour day). The sidebar slot is not currently contested by any of our products — it is owned by Twitch, YouTube, Spotify, and cable-style news.

The proposal: give the portfolio one shared companion surface — the Rail — into which every V-product publishes an ambient channel. Full immersion remains a door; it stops being the only door.

Why broadcast media owns the slot (and games don't)#

Four properties, all of which are design requirements for any Rail channel:

1. Progresses without you. A stream doesn't pause when you look away; a game stalls the instant you stop inputting. Channels must be backed by something persistent — a league, a simulation, a case unfolding, a live feed — that moves on its own.
2. Glance-legible. A scoreboard or a chat can be reconstructed in two seconds. Every channel must render its "what's happening now" in a 2-second-readable tile.
3. Absence is free. Miss an hour, miss a day — nothing punishes you. No appointment mechanics, no decay, no FOMO traps. (This aligns exactly with V1's existing humane-streaks stance in Arete — done|partial|skip|decline|miss — which the Rail adopts portfolio-wide.)
4. It's company. Chat scroll, crowd noise, a voice — co-presence without obligation. The mood function is a first-class goal, not a side effect.

The strategic asymmetry#

The portfolio already built most of the broadcast organs, aimed at other purposes: V4's commentary/director pipeline, V5's broadcast-director and highlight ranker, V7's spectator-funnel recommendation, V2's esports-flagship recommendation, and — the largest single asset — Aphrodite's complete live streaming stack (seventeen deployable services: ingest, broadcast, viewing, chat, payments, delivery, notifications; see §10). The ambient tier is not a tenth product to build from scratch. It is a re-aiming of spectator substrates that already exist and that the consolidation audit found duplicated across Vs anyway.

2. Decisions already taken (from the 2026-07-16 riff)#

These were explicitly settled and are treated as fixed inputs by the rest of this document:

3. What the Rail is#

A persistent companion surface that lives beside work, not in front of it.

Product definition. A native desktop companion shell (D7): its own app, sized and shaped for the classic two-thirds/one-third desktop split, that hosts: a stack of channel presence tiles, one merged timeline of drip events, an arbitrated audio lane, an arbitrated video lane, and a micro-act surface. It choreographs all of these against the user's dayparts (§6). Window behavior is the product: launches narrow, docks to a screen edge, stays pinned across restarts, offers always-on-top, and keeps a tray/menu-bar presence (tile glances + audio control) even when the panel is closed. The UI inside the shell is web-tech and shared with the P1 browser fallback; the shell itself is Tauri (Rust — decided with D7, per the stack principles) — a plain web app cannot deliver tray, always-on-top, or reliable pinning, which is precisely why D7 rejects the PWA-only path.

Why it rides V1's substrate. (Ownership: the Rail is its own product, V10 — D8. These three structural facts make V1 the substrate underneath it, not its owner:)

1. V1 is the only V that already spans web, PWA, mobile, and watch — the Rail's surface family (§3.1).
2. V1's review already proposed the composed daily ritual (the Thread, review §4.1/§5.1: wake → today's practice → briefing → …). The Thread is the Rail's morning daypart, verbatim.
3. The consolidation work already assigns V1 the account/entitlement graph. The Rail is an identity-heavy surface (channel subscriptions, loudness grants, ring opt-ins) and should sit on that graph, not beside it.

The Rail is therefore also V1's diffuse-positioning answer: the review's core criticism was "a cathedral with six entrances and no nave." The Rail is the nave — and V1's own domains become house channels inside it (§7.1). With D8 the nave is its own building — V10 — rather than a seventh entrance carved into V1: V1's domains tenant the venue like every other V, and V1 sheds surface area instead of gaining it.

3.1 Surfaces#

App-store note: the mobile/watch satellites can never carry the Aphrodite ring (§9); this is a per-surface channel-availability matrix from day one, not an afterthought.

4. The channel contract#

Every V publishes a channel by implementing five primitives. The Rail owns composition, batching, arbitration, and policy; channels own content.

4.1 Presence tile#

The 2-second-legible "now" render: an evidence board's state, your ghost's league record, tonight's sky event, cards due, realm population. Constraints: silent; no autoplay motion by default (peripheral motion is involuntary attention capture); must render meaningfully from cold cache; must degrade to text-only (required for discretion mode, §9.3, and the watch surface).

4.2 Drip events#

Discrete, timestamped happenings: a witness statement released, your ghost upset a higher seed, a creator shipped to your realm, a claim was verified live. Channels cannot interrupt — they emit events into one merged timeline; the Rail decides batching, ordering, and delivery timing as a matter of daypart policy. Calm stays enforceable only because batching is a Rail policy, not a channel courtesy.

4.3 Micro-acts#

Interactions of ≤30 seconds, executable inside the Rail without launching the product: pin a suspect, answer two review cards, tweak the ghost's tendencies, approve a realm collab, do one breath with Tara. Micro-acts are what make the Rail part of the day rather than decoration — and they are the top of the dive funnel (§8).

4.4 Dual faces#

Every channel declares a spectator face and a player face on the same tile (decision D2). Spectator: featured league matches, realm channel-surf, today's public case. Player: your ghost, your realm's overnight economy, your evidence board. The Rail picks the face from entitlement state; a spectator can watch for months before the first dive, and that is a success state, not a failed conversion.

4.5 Live moments#

The earned interruptions: an accusation window closing, the ghost final, a concert going live, a breaking Veritas event. Channels request elevation; the user's standing per-channel grant (§5) decides. Requests are typed (scheduled vs. emergent) and rate-limited per channel per day by the Rail.

4.6 Contract sketch (illustrative)#

// Shape sketch for the concept doc — not an implementation.
interface ChannelManifest {
  id: string; // "v8.case-files"
  ring: 'wellness' | 'games' | 'adult'; // §9
  surfaces: SurfaceAvailability; // web/mobile/watch/tv availability
  faces: { spectator: TileSpec; player?: TileSpec };
  lanes?: { audio?: AudioSourceSpec; video?: VideoSourceSpec };
  microActs: MicroActSpec[]; // each with an expected-duration budget
  liveMoments: LiveMomentSpec[]; // typed, rate-limit class, daypart hints
  daypartHints: DaypartAffinity; // where this channel naturally lives
}

5. Loudness: calm by default, granted never taken#

The single mechanism that reconciles "respect attention" (headline) with "engaged players want to be interrupted" (real utility):

Per-channel loudness grant, set by the user:

Effective loudness = min(daypart ceiling, channel grant), with explicit user-pinned exceptions ("always break through for V2 finals"). Deep-work dayparts clamp the ceiling to 1 (audio lane exempt); breaks raise it; evenings open it. The default grant for every new channel is 1.

Rail-level invariants, non-negotiable across all channels and rings:

• No channel may punish absence (no streak loss, no decay, no expiring rewards keyed to attention).
• No autoplay motion or sound at loudness ≤2.
• Elevation requests are budgeted per channel per day; unspent budgets do not accumulate.
• Success metrics for the Rail are docked-hours and micro-act quality, never interrupts-clicked (§12 — measurement is where calm products quietly die).

6. Lanes and dayparts#

6.1 The audio lane#

Exactly one holder at a time — the Rail is the mixing desk. Audio is the most powerful work-companion medium and the easiest to ruin by letting every channel talk. Candidate holders: V3 adaptive work-music, V5's diegetic period radio, match commentary (V2/V4), Ori's voice (V6), the Veritas audio briefing (already proposed in the V1 review, §5.8). Handoffs are explicit (user action or daypart transition), never a channel grabbing the mic.

6.2 The video lane#

One primary holder at a time, same arbitration. Three content classes:

1. First-party live — Veritas produced coverage; V3 Stage concerts; auto-directed game broadcast (below).
2. Overlay / second-screen — the video plays wherever it plays (even YouTube or broadcast TV); the Rail runs the synced companion layer, e.g. Veritas live claim-verification ticker against an external stream. Lower rights/infra burden; ships before first-party (§11).
3. Docked third-party (decision D5) — the user's own YouTube/Twitch/etc. lives in the lane. We host the embed, we do not moderate the content; our channels grow alongside it. This is how the Rail wins the slot geometry on day one instead of fighting the incumbent for a second sliver of screen.

Attention-aware director profiles. The auto-director (V4/V5 pipelines) gets two cut grammars: ambient cut (wide shots, slow cuts, low motion, commentary murmur or off) for deep-work dayparts, and highlight cut (fast cuts, replays, loud commentary) for breaks and evenings. Same pipeline, two profiles, switched by daypart or by the user. This is the concrete mechanism that makes a video lane compatible with calm-by-default.

24/7 auto-directed channels. Because the director and commentary organs are software, first-party game channels run without human streamers — a league that plays all day, a world that is simply on. No talent costs, no schedule gaps; the marginal cost is compute and delivery.

6.3 Dayparts#

The Rail's choreography spine. Defaults (user-tunable, calendar-aware later):

Absence remains free in every daypart: dayparts gate what may surface, never what the user owes.

7. The channels, V by V#

Fit legend — born-ambient (the ambient form may be the better product), broadcast-ambient (ambient via its broadcast of itself), audio-ambient, immersion-first (companion tier feeds a destination product).

7.1 V1 Oshun — the house channels (the venue itself is V10 — D8)#

Fit: born-ambient. The venue is V10 (D8); V1's six domains are themselves channels: the Thread as the composed morning daypart (review §4.1); Veritas Live as the flagship news channel — live events with claims extracted and verified in the ticker as they are spoken, provenance/C2PA receipts on the feed, in both first-party and overlay modes ("live news you can leave on during work, with a trust meter" — differentiated against cable and YouTube news, and it serves the front-door audience that will never touch a game); Nyx as sky-and-awe ambience (tonight's sky events, live sky cams, aurora alerts — awe as a wind-down mood channel); Tara as a presence micro-channel (an hourly bell, one breath, never a guilt mechanic); Arete as a gentle progress tile carrying its humane-streak semantics into the Rail's invariants; Nisaba/Metis folding into the V9 channel (§7.9). Substrate: V1's four-surface stack; @oshun/content-signing for feed provenance (the audit's chosen canonical C2PA signer).

7.2 V2 Fighting — Ghost Dojo League#

Fit: broadcast-ambient; the announced destination channel (not a pilot). Your trained ghost fights a persistent league all day while you work. Tile: record, rank, next bout. Drip: results, upsets, rival callouts. Micro-acts: adjust tendencies, scout the next opponent, call the outcome (prediction as points/bragging rights only — anything wager-shaped trips gambling policy, §12). Video lane: auto-commentated bouts, ambient cut during work. Live moment: your ghost's final. Spectator face: the featured league — salty-bet psychology without the betting. This is fantasy-sports psychology bolted to a fighting game and is probably the portfolio's most magnetic ambient concept — but the V2 review found zero cooked content, and a ghost league without real fighters and a real ladder reads as a demo. Announce it as the destination; do not gate the Rail on it. Substrate: libs/shakti/fighting-ruleset-bridge (the audit's only working cross-domain game bridge); Ghost Dojo already recommended by the V2 review.

7.3 V3 Lilith — Motion (audio) + Stage (video)#

Fit: audio-ambient (Motion), broadcast-ambient (V3 Stage). Both pilots. Motion: adaptive work-music holding the audio lane, with movement-break micro-acts (stretch, posture, a 90-second dance break) and class schedules as live moments; the review's passthrough Mat Mode is the immersion end of this funnel. V3 Stage: music television — music videos, live concerts, stage performances in the video lane. Music television is the original ambient TV; it pairs natively with the audio lane (the video is optional visual garnish on music you were already going to play), needs no game to ship, and makes the Rail worth docking on day one for someone who ignores every other channel. Concerts are live moments and the natural first watch-party content. Substrate: V3's music/stage stack plus Calliope's cited stage specification. The technical channel id is v3.stage; legacy @oshun/tenant-saraswati-stage implementation names remain distinct from the unrelated @saraswati/* industrial-technology domain. The audited relationship and render gaps are recorded in RA8_V3_STAGE_CHANNEL_INTEGRATION.md.

7.4 V4 Tactical — the Match Channel#

Fit: broadcast-ambient (post-launch). A 24/7 auto-directed, auto-commentated match channel — the commentary/director pipeline is V4's standout artifact and this is its second job. Ambient cut during work; highlight cut on breaks; esports events as live moments and watch parties. Player face: your squad's persistent campaign ticking between your sessions. Waits on the game being live; the pipeline work (director profiles, §6.2) starts earlier because Veritas and V3 Stage exercise the same lane. Substrate: V4 commentary/broadcast pipeline (real per the 2026-07-14 audit, with its consent/bias gates now evidence-gated fail-closed — those gates are a feature for a channel that broadcasts publicly).

7.5 V5 Open-world — the Period Channel#

Fit: born-ambient in concept; gated on the world shipping. The strongest expression of "progresses without you" in the portfolio: a persistent simulated period world that is simply on. Diegetic radio in the audio lane; the world's daily newspaper as drip (procedurally real — generated from what actually happened in the simulation); slow-TV world windows in the video lane (streets, harbor, trains — ambient cut by construction); and the Mind Palace detective layer as a slow-drip case, which the reviews already connect to V8's engine. The V5 review's verdict (the real product inside is the detective layer) makes the Rail its cheapest proof: newspaper + radio + one district window is a channel long before the full open world is a game. Substrate: V5 broadcast-director + highlight ranker + persona/daily systems (real per the 2026-07-14 audit); V5's 17 substrate services.

7.6 V6 Egbe — Ori Presence#

Fit: born-ambient in the purest sense — presence is the product. Ori is not a tile; it is the Rail's optional companion layer. A greeting woven into the Thread; consent-gated awareness of your dayparts; remembrance and ritual moments surfacing gently on meaningful dates; a voice that can hold the audio lane briefly; and — the role with portfolio-wide leverage — Ori as the Rail's narrator/concierge: "your ghost won both bouts while you were in that meeting; the case's third witness contradicts the first — want the 40-second recap?" That single behavior stitches every other channel into a relationship, which is exactly the keystone/loyalty role the reviews assigned Ori. V6's consent architecture becomes the Rail's governance substrate for anything presence-shaped. Honest gate: the V6 review flagged LLM wiring and inference economics as unproven — Ori Presence ships behind a believability beta, and the Rail must be whole without it. Substrate: V6 consent/grief architecture; Ori continuity guarantees (second-pass review).

7.7 V7 Mawu — Realm Surf + My Realm#

Fit: broadcast-ambient (spectator) + born-ambient (creator). Spectator face: Realm Surf — channel-surfing live windows into UGC realms (the review's spectator funnel, made literal). Creator/player face: My Realm — your realm's economy ticking overnight (visitors, sales, sim events) with creator micro-acts: approve a collab, adjust a price, greenlight a patch. For creators this is a utility companion (the shopkeeper's back-office glance), which may be the strongest daily-return loop in the portfolio. Gated on the network runtime the review flagged; the creator-face tile can precede public realm broadcast. Substrate: libs/maya/crucible-* including crucible-live; Mawu economy sim (real per the 2026-07-14 audit); vendor-wired safety precedes any executable UGC in the video lane.

7.8 V8 Ariadne — Case Files (the anchor pilot)#

Fit: born-ambient — the dripped case is a better product than the sit-down mystery, and it is Wordle-shaped for virality. Morning: the case opens (victim, setting, cast). Across the workday: director-paced clue and witness releases as drip. Tile: the evidence board's state. Micro-acts: pin a suspect, mark a contradiction, take a proof-tree minimal hint (second-pass review concept). Evening live moment: the accusation window — then the reveal with the provable-fairness receipt, which is the moat: the only daily deduction game that can prove the case was solvable from the released clues. Social face: compare deductions with friends after the reveal, never before. Weekend: bigger cases. This channel is the "standalone daily-case app" the V8 review asked for, with better daypart mechanics than a standalone could justify — and it routes around the review's GA-blocker (V5 world data) because a daily text-and-tableau case needs the case engine, not the open world. Substrate: libs/v8/case-* + libs/yemaya/case-* (compiler, director, engine, eval — real/SOTA per the 2026-07-14 audit: CSP+DPLL uniqueness proofs, solve-first generation, Ed25519 C2PA signing).

7.9 V9 Metis — Wonder & Recall#

Fit: born-ambient — FSRS is literally an ambient-cadence algorithm. The Daily Wonder drops in the morning brief. Spaced-repetition micro-reviews surface as break filler — the scheduler already wants to distribute 40-second interactions across your day; the Rail is the surface that cadence has been missing. Kernel drops as events; fading-stars as a gentle tile; the weekly Wonder Report (second-pass review) as a wind-down moment. The ≤5s first-truth SLO from the review is exactly the glance constraint. The V9 verdict was "substrates real, consumer app missing" — the Rail is the missing consumer app, at a fraction of a standalone app's cost. Substrate: libs/v9/* (aletheia, atlas, chiron, theia, prometheus, lesson-gates…) with real FSRS/SM-2/IRT per the 2026-07-14 audit; V1's Nisaba/Metis rooms fold into this channel rather than duplicating it.

7.10 Aphrodite — substrate donor and walled tenant#

Fit: broadcast-ambient content; portfolio-critical infrastructure. Two distinct roles, deliberately separated: (a) Substrate. Aphrodite is the only fully-built live-streaming stack in the monorepo — seventeen deployable services covering ingest, broadcast, viewing, chat, payments, delivery, notifications, plus mobile shells (systems/app-aphrodite.md). The video lane (§6.2) should be built by generalizing these organs rather than building a parallel stack — the same consolidation move the platform audit keeps recommending, applied to video. Veritas live, V3 Stage, and the game channels all become tenants of the stack Aphrodite donates. (b) Content ring. An adult live-streaming channel can exist only inside the hard-walled adult ring (§9): explicit opt-in behind age verification, separate brand, web-only (app-store policy excludes it from mobile/watch satellites), daypart-locked to private/wind-down contexts, and discretion mode as a hard requirement. Aphrodite's own consent/aftercare/body-dignity libraries indicate the domain already takes this architecture seriously; the Rail inherits and enforces it at the surface.

7.11 Fit summary#

8. Audiences and the funnel#

The Rail serves both audiences with one surface because every channel carries two faces (§4.4). The funnel is: glance → micro-act → live moment → dive, and it is deliberately shallow at the top — sports psychology, not casino psychology. You leave the league on all day; the final earns your full screen; occasionally you buy a ticket to the stadium.

Two funnel rules:

• Spectating is a success state. The front-door audience (D2) may watch for months and never dive. They still hold the slot, still build the habit, still join watch parties, still tell friends. Do not measure them as failed conversions.
• Every tile deep-links with context. A dive launched from the Rail lands inside the moment that prompted it (the bout, the accusation, the concert), never on a main menu.

9. Brand rings and policy walls#

The reviews' two-brand question (wellness vs. games) becomes a three-ring model at the Rail:

9.1 The adult ring wall#

Never in the default channel directory or recommendations; explicit opt-in behind age verification (Aphrodite's own age-verification service); separate brand identity end-to-end; web-only (no mobile/watch/TV satellites); eligible only in private/wind-down dayparts; excluded from Ori narration, watch parties, and all cross-channel promotion.

9.2 Payments and gambling adjacency#

Prediction micro-acts (V2/V4) are points-and-bragging-rights only. Anything that lets value in one side and out the other is out of scope for this concept and would require its own review.

9.3 Discretion mode — a Rail-level feature, not an adult-ring feature#

A glanceable sidebar is the opposite of private, and work screens get shared. Requirements: one-keystroke instant hide; per-channel text-only tile option; and screen-share detection auto-engaging discretion (personal tiles — health, Ori, adult ring, even the case you're solving — collapse to neutral text the moment the screen is shared). This protects the wellness ring as much as the adult one.

10. Substrate reuse map (what already exists, as of 2026-07-14 audits)#

The duplication census (7 matchmaking impls, 5 rollback, ~6 battle-pass, 3 voice/SFU, 2 C2PA signers) is the cautionary tale: the Rail must be the consolidation, not an eleventh duplicate. Channel contract, streaming substrate, and signer are each built once.

11. Pilots and sequencing#

Phase A — the Rail exists (P0). The native desktop shell (D7 — docked window, always-on-top, tray presence, V1 identity underneath): tiles, timeline, loudness grants, dayparts, audio lane, video lane with third-party docking (D5 — this alone makes the Rail worth docking on day one), discretion mode. Four pilot channels:

1. V8 Case Files — the anchor; proves drip + micro-acts + deadline + the evening live moment.
2. V9 Wonder & Recall — proves the learning cadence; V9's consumer debut.
3. V3 Motion — proves audio-lane arbitration; the retention floor.
4. V3 Stage — proves first-party video with zero game dependency.

Exit test for Phase A, in order of honesty: does the author of this concept actually leave it docked next to Claude Code for a full week?

Phase B — the flagship and the voice (P1). Veritas Live, overlay/second-screen mode first (rights-light), then first-party produced coverage. Begin Ori Presence believability beta (narrator/concierge behind a flag). Extract the Aphrodite streaming substrate properly (the Phase A video lane may run on a thinner slice).

Phase C — the game broadcast wave (P2). As games ship: Ghost Dojo League (the announced destination — market it from Phase A, ship it when V2 has cooked content), Realm Surf / My Realm (creator tile first), Period Channel (newspaper + radio before world windows), Match Channel post-V4-launch. Director cut profiles land here but are developed against Phase B's Veritas/V3 Stage streams.

Phase D — rings and satellites (P3). Mobile/watch satellites; TV/watch-party surface; the Aphrodite content ring behind its full wall; calendar-aware dayparts.

12. Risks and open questions#

1. The engagement-metric trap. Every incentive will push toward measuring interrupts-clicked. Lock the metrics now: docked-hours, micro-act completion quality, voluntary-return rate, and a calm SLO (elevations per user-day stays under a ceiling). If the Rail starts optimizing attention capture it becomes the thing it was built to replace.
2. Veritas live rights and cost. First-party live coverage is a rights/ops business, not a tech feature. Overlay mode first is the hedge; first-party scope needs its own proposal.
3. Content cook remains the portfolio disease. The reviews' recurring finding (plans ahead of cooked content) applies: Phase A channels were chosen specifically because their engines are real, but V8 needs a case pipeline running daily and V3 Stage needs a rendered-media cook plus content operations. A channel that drips nothing is worse than no channel.
4. Notification fatigue by a thousand grants. Ten channels at loudness 2 is still a noisy timeline. The Rail may need a global drip budget on top of per-channel limits.
5. Moderation surface of docked third-party video. We host the embed, we don't moderate the content — but the boundary needs legal review, especially once watch parties add our chat next to their stream.
6. Ori economics. The V6 review's inference-cost concern applies double to an always-present narrator. Believability beta + cost ceiling before any default-on behavior.
7. Gambling adjacency of prediction mechanics (§9.2) — keep points-only, revisit never lightly.
8. Privacy of glanceable surfaces. Discretion mode (§9.3) is load-bearing for trust in the wellness brand; treat screen-share auto-hide as P0, not polish.
9. Resolved 2026-07-17 — naming. "The Rail" is a working name; "the Thread" names the morning ritual inside it (per the V1 review). The venue channel is technically V3 Stage, id v3.stage; historical @oshun/tenant-saraswati-stage/Unreal/data names remain in V3, while @saraswati/* unambiguously remains the EV/battery/manufacturing technology domain. Final public branding stays HG-7.
10. Resolved 2026-07-16 → D7. The Rail ships as its own native shell on V1's identity, with tray presence, always-on-top, and docked window management as P0 requirements — the docked geometry is a different product posture than V1's tabs, and those window behaviors are unreachable from inside a web app or plain PWA. Wrapper decided the same day: Tauri (Rust core, smaller resident footprint than Electron, consistent with the stack principles and with the portfolio's existing Rust investment).
11. Resolved 2026-07-16 → D8. The Rail is V10 — its own V-series product, with V1 remaining the identity substrate (D7 unchanged). See D8 for the recorded rationale; V10 must be registered in scripts/audit/capability-truth-registry.json so the generated V_SERIES.md picks it up (checklist task X-8).

13. Closing#

The V-series keeps being reviewed as nine destinations of varying readiness. This document argues the portfolio's fastest route to daily relevance is not finishing nine cathedrals — it is claiming the sidebar slot that modern attention actually lives in, with one calm, dual-audience companion surface that each V feeds at whatever fidelity it can honestly sustain today: a case that drips, a scheduler that recalls, music that plays, a stage that streams, a league that fights while you work, a world that goes on without you, a companion who noticed. Full immersion stays on the menu. It just stops being the price of admission.