PRD-034 — The Known Universe (the outward arc of /explore)

Status · Shipped (v0.8.0, 2026-08-27) — /explore extended across the observable universe (8 scale shells, scale picker, time playback, message-object trajectories, Worlds IA split, cosmology tab). See CHANGELOG [0.8.0]. Sources · #259 (/cosmos) · #258 (scale picker, shipped) · #410 (message-objects, shipped) · RFC-032 (/explore v2) · PRD-030 (Known Universe principles) · ADR-073 (deferred surface work) Audiences · curious learner, STEM student (PA §audiences) Promises · real physics, educational at every level, honest-when-schematic (PA §promises) Principles · physics first, honest schematic badging, prototype is ground truth (PA §principles)

Why this is a PRD. "Zoom out to the edge of the observable universe" sounds like one route (#259). It is not. Doing it honestly forces three interlocking product decisions at once: (1) how the scale-navigation UX works across eight tiers instead of four; (2) whether the four existing shells — especially the Milky Way, which today has two clickable objects and zero science lenses — are good enough to sit inside a coherent whole, or must be completed first; and (3) how far the learning surface stretches, because a known-universe atlas that can't explain the Big Bang, nucleosynthesis, the distance ladder, or dark matter is a picture without a story. These three are not separable — a bare Milky Way inside a gorgeous cosmic-web zoom would read as broken, and new cosmology tiers with no science articles behind them would violate the "educational at every level" promise. They need one agreed frame before any code lands.

T+0. The camera sits over the Sun. The user drags the scale outward: the planets collapse to a point, the nearest stars bloom and thin, the Milky Way's arms wheel past, the Local Group's galaxies drift into view — and then, where the app currently hard-stops, it keeps going: the Local Sheet, the Virgo Cluster with M87 at its heart, the Laniakea supercluster flowing toward the Great Attractor, and finally the cosmic web itself, filaments and voids, the largest honest structure we can draw. At every tier a science lens is one tap away — "how do we know how far?", "what is the Hubble flow?", "where is the dark matter?" — and every schematic wears its "not to scale" badge without apology. One continuous outward breath, from a pad on Earth to the edge of the observable universe.

North star

The bar is a planetarium's "Powers of Ten" show, made interactive and honest — Charles-and-Ray-Eames pacing with a modern astrophysical conscience. Not a tech demo of a big number; a guided, legible, truthful ascent through scale where every rung teaches something and nothing pretends to a precision it doesn't have. It measures itself against the Cosmic-Eye / Digital Universe (AMNH) lineage, the way /fly measures itself against a live launch broadcast.

The problem — three gaps, one incoherence

A full gap-audit (3 parallel recon passes, 2026-08-08) found:

1 · The arc dead-ends at the Local Group. scale-shell-controller.ts has exactly four rungs; context-graph.ts sets local-group.parent = null ("outermost for now; the wider universe arrives later"); the wheel handler at the Local Group ceiling just clamps and stops. There is essentially zero cosmology data past the Local Group — a repo-wide search for Virgo / Laniakea / supercluster / cosmic-web / Great Attractor / redshift / Hubble-flow returns nothing but a constellation string and one mislabelled M87. 34 Virgo galaxies already sit in deep-sky-objects.json with dist_ly: null.

2 · The existing shells are wildly uneven. The solar system is dense (planets, moons, small bodies, belts, 18 iconic trajectories, message-objects, culture doors, dozens of science articles, a full science-lens layer set). By contrast:

• Milky Way shell: 2 clickable objects (Sun pin, Sag A*), no galactic bar (the science article's see_in_app literally mentions a bar the scene doesn't draw), no globular clusters, no scale HUD (it's suppressed), 1 science article, zero lens chips, and the Sun's panel renders mostly-empty cells.
• Stellar Neighbourhood: star data is position+magnitude+colour only (no proper motion → no PM lens possible), 52/62 named stars have no photo, every star has a single Wikipedia link, 97/141 deep-sky glints are dead clicks, science coverage is thin.
• Local Group: two partially-duplicated data files, a panel with no learn-more link, a honesty badge with no learn-more link.

3 · The learning surface stops at the Local Group, conceptually. 162 science articles exist across 14 tabs — but on cosmology the cupboard is nearly bare. Absent entirely: Big Bang, nucleosynthesis (primordial + stellar), the role of carbon / CNO, supernovae, the cosmic distance ladder, Cepheids/standard candles, redshift & Hubble's law, dark energy, the CMB, inflation, large-scale structure. Dark matter and stellar evolution exist only as incidental mentions inside observational articles, never as explainers. And the science lens is uneven: it's rich on the solar system, /fly, and the surfaces, but the Milky Way and Local Group shells have no lens at all, and several teaching-grade chips (DEEP-SKY, CONSTELLATIONS, PATHS) live as always-on "service" toggles outside the lens rather than inside it.

The incoherence: you cannot ship a beautiful cosmic-web zoom on top of a Milky Way with two dots and no lens. The whole arc has to read as one coherent, educational, honest instrument. So the epic is not "add /cosmos." It is complete and level the arc, then extend it, and thread a consistent learning layer through the whole thing.

The three pillars

Pillar 1 — Extend the arc (the new outward tiers)

Add four tiers past the Local Group, for an eight-rung ladder:

Tier	Scale	What's there	"You are here" anchor
Local Group (exists)	~3 Mpc	MW + M31 + M33 + ~30 dwarfs	Milky Way
Local Sheet / Volume	~10 Mpc	M81 group, Cen A, Sculptor, Maffei, IC 342	Local Group blob
Virgo Supercluster	~33 Mpc	Virgo Cluster (M87 BCG, ~1300 galaxies), Fornax, our supercluster plane	Local Sheet
Laniakea	~160 Mpc	supercluster flow-basin, Great Attractor (Norma), Coma, Perseus-Pisces	Virgo Cluster
Cosmic Web	~1–14 Gpc	filaments, walls, voids, → the observable-universe / CMB shell	Laniakea

Each new tier = a scene module (*-scene.ts), a schematic data file (supergalactic-coordinate positions, honest + provenanced), a panel, an honesty badge, a scale-picker rung, and crossing functions — following the exact pattern the Local Group shell already establishes. Every tier is explicitly schematic (real catalogue distances, illustrative positions) and badged as such.

Pillar 2 — Complete + level the existing shells

Before/with the new tiers, raise the thin shells to a coherent bar. Candidate work (to be prioritised in the RFC):

• Milky Way: add the galactic bar, a globular-cluster layer, the Magellanic Clouds at the disc edge, arm interactivity (click an arm → panel/article), a scale HUD, and — critically — a science lens (rotation curve, dark-matter halo, stellar populations, the galactic centre).
• Neighbourhood: enrich star panels (photos, tiered links beyond Wikipedia), make dead deep-sky glints either open something or dim out, add a proper-motion lens (needs a data-schema extension to carry pm/parallax), tie clusters to their member stars.
• Local Group: unify the two data files, add learn-more links to the panel + badge, enrich galaxy fields.

Pillar 3 — Expand the learning + science-lens surface

• New cosmology articles (the "key things to know about the known universe"): Big Bang, primordial + stellar nucleosynthesis, the role of carbon / CNO cycle + carbon-based life, stellar evolution (full life cycle), supernovae + standard candles, the cosmic distance ladder (parallax → Cepheids → SN Ia), redshift & Hubble's law, dark matter (dedicated explainer), dark energy, the CMB, large-scale structure / cosmic web. Likely a new cosmology science tab (or a big observation expansion), full article shape (formula, diagram, narrative, tiered links, see_in_app into the new tiers), all 14 locales.
• A science lens on every shell. Wire ScienceLayersPanel (or the inline lens-chip pattern) into the Milky Way + Local Group + new cosmology shells so each has real teaching overlays, not just a badge.
• Reclassify chips: service vs teaching. Audit the current chips; move teaching-grade ones (DEEP-SKY, CONSTELLATIONS, and the science overlays that today sit outside the lens) into the science lens, keeping pure-utility visibility toggles (PLANETS/DWARFS/COMETS/INTERSTELLAR) as always-on service chips. Make the lens the single consistent home for "this teaches you something."

UX — reuse the patterns, pick a scale-nav direction

This epic inherits the established /explore UX language and must honour it: the #258 scale picker + breadcrumb, the mobile drawer pattern (MobileDrawerGroup), the culture-door cards, the honest schematic badge (PRD-030 principle 2), and the critical-review cadence we've been running (mock-gate → build in slices → science-reviewer fact-check → i18n at the end → e2e on desktop+mobile).

Scale-nav direction — DECIDED: A · Extended ladder rail (operator, 2026-08-08). Grow the shipped #258 ExploreScalePicker from 4 rungs to 8; keep it the single consistent scale control — discrete, tappable, and byte-consistent with what already shipped. Mobile follows the same collapse-to-"Scale ▾"-chip → popover pattern from #258 (the 8-rung popover replaces the 4-rung one). Rejected: B (continuous slider — precise-tier targeting is hard on touch) and C (hybrid spectrum — largest footprint) — kept on file as fallbacks if the 8-rung rail proves too tall in practice.

Architecture — DECIDED: extend /explore, not a separate route (operator, 2026-08-08). The new tiers are added as more shells on the existing ladder — one coherent instrument, reusing the crossing / picker / breadcrumb / honesty-badge infra. The "separate /cosmos route" framing of #259 is retired; #259 becomes "extend /explore's shell ladder to the observable universe."

Phasing (proposed)

0. This PRD + mock-gate — pick the scale-nav direction; agree the pillar sequencing.
1. RFC — freeze the shell contract (context-graph extension, supergalactic coords, data schemas, the science-lens-per-shell wiring, chip reclassification).
2. Level the Milky Way (highest-leverage completeness win + first new lens).
3. Extend the ladder one tier at a time (Local Sheet → Virgo → Laniakea → Cosmic Web), each with data + scene + panel + badge + science article(s), science-reviewed + 14-locale i18n.
4. Learning-surface expansion interleaved (each tier ships with its backing articles; the foundational cosmology articles — Big Bang, distance ladder, carbon — can lead).
5. Only then the grand-tour narration (#411) rides on top of a now-coherent, content-rich arc.

Non-goals (this epic)

• Not real-time cosmological simulation or N-body; schematic, catalogue-backed, honestly badged.
• Not photorealistic to-scale rendering of 10²⁷ m (impossible + dishonest) — legible schematic + honest scale readout.
• Not a redesign of the solar-system shell (it's the reference bar, not the subject).
• Not #411 narration yet — it's the capstone after the surfaces are complete (operator's explicit sequencing).

Decisions (2026-08-08)

1. Scale-nav direction — ✅ A · extended ladder rail (grow the #258 picker to 8 rungs).
2. Route — ✅ Extend /explore (retire the separate-/cosmos-route framing).
3. Pillar sequencing — ✅ Level the Milky Way first, then extend outward one tier at a time.
4. Science tab shape — ⛶ OPEN: a new cosmology tab vs a large observation expansion (decide in the RFC).

Next step

An RFC to freeze the contract before code: the context-graph extension (8 shells), the supergalactic-coordinate system + schematic data schemas, the Milky-Way-leveling scope (bar, globulars, scale HUD, first science lens), the science-lens-per-shell wiring, and the chip service-vs-teaching reclassification. Milestone 1 after the RFC is leveling the Milky Way (with its own mock-gate for the enrichment look).

---

PRD-034 · draft · 2026-08-08 — scoping the outward arc. Audit detail in the session recon; mocks on the operator's Desktop.