feat(cosmos): constellation star maps + 11-card encyclopedia wave (v2)

docs/superpowers/specs/2026-07-13-cosmos-platform-expansion.md

@@ -0,0 +1,62 @@
+# Starlight Cosmos — Platform Expansion Spec
+
+> 2026-07-13 · Operational-tier (site surfaces + plan doc). Extends `2026-06-11-starlight-cosmos-design.md`.
+> Built on SIP. Captures Frank's platform vision (encyclopedia · visual exploration · story + research agents · multi-surface apps · monetization · network visualization) as a phased, falsifiable plan.
+
+---
+
+## 1. What shipped this session (v2 — Constellations & Spacecraft wave)
+
+- **Encyclopedia +11 cards** (19 → 30): 4 constellations (Orion, Ursa Major, Crux, Cassiopeia), 3 named stars (Betelgeuse, Sirius, Proxima Centauri), 2 spacecraft (Voyager 1, ISS), 2 concepts (Fusion on Earth, What Stars Teach — the philosophy layer, kept honest by physics).
+- **Two new card kinds**: `constellation`, `spacecraft` — the registry now spans 12 kinds.
+- **First visual-exploration surface**: `/cosmos/constellations` — deterministic SVG star maps projected from real J2000 RA/Dec, zero client JS, three reading layers per chart (science / myth / navigation).
+- **Dual-track prompts**: every new card carries a systems prompt, a research trail, and a story seed — the same substrate serves scientific exploration and fiction worldbuilding without mixing the two.
+
+## 2. The thesis, restated
+
+The cosmos surfaces are the public, beautiful proof of the SIS claim: knowledge as agent-consumable substrate. Humans get an encyclopedia with taste; agents get the same registry as MCP resources; the prompts are the bridge between them. Everything below scales that one loop — never a second, disconnected product.
+
+## 3. Vision → workstreams
+
+| # | Workstream | What it is | Status |
+|---|---|---|---|
+| W1 | **Encyclopedia scale-out** | 30 → 100+ cards: planets complete, key moons, more constellations (zodiac, southern sky), spacecraft fleet (Hubble, Cassini, New Horizons, Artemis), stellar lifecycle chain, philosophy-of-science cards | Continuous, every session |
+| W2 | **Visual exploration** | Constellations page (✅) → spacecraft schematic explorer → 3D starfield (r3f, HYG subset, per prior spec S+4) → orbital mechanics playground | Phase-gated |
+| W3 | **Agents on the substrate** | Cosmos MCP v0.1 in `starlight-cosmos-engine` (cards as `cosmos://cards/{slug}` resources, prompts as MCP prompts, NASA fetchers as tools) → story-forge + research-companion agent presets consuming it | S+2 per prior spec — next code session in cosmos-engine |
+| W4 | **Site agent** | "Ask the Cosmos" — an embedded agent on starlightintelligence.org answering from the card corpus (Vercel AI SDK, RAG over registry + markdown, cite cards). Gate: W3 first, so the agent consumes the same MCP the public gets | After W3 |
+| W5 | **Multi-surface** | Expo/React Native app (offline card library + tonight's-sky view), desktop (Tauri preferred over Electron for footprint) — one content pipeline, three renderers | After 100+ cards (content moat first, apps second) |
+| W6 | **Network visualization** | Starlight Network view: SIP nodes, protocol flows, attestation chains rendered like the knowledge-tree graph. Honest framing: visualization of the real SIS/SIP topology — no blockchain claims until an actual chain integration exists | Design doc first |
+| W7 | **Monetization** | See §4 | Decision needed |
+
+## 4. Monetization — honest options, one recommendation
+
+Constraints: the encyclopedia must stay free (it's the distribution + SEO engine and the proof of the thesis); no paywall in front of knowledge cards; OSS substrate stays MIT.
+
+| Model | What's paid | Risk |
+|---|---|---|
+| A. Cosmos Pro subscription | Site agent (W4) with generous free tier, saved exploration trails, exports | Low — standard SaaS on top of free content |
+| B. Templates & packs | Story-universe starter kits, research-workflow templates, agent presets — one-time purchases | Low — matches existing FrankX template motion |
+| C. Community lifetime | Founding-member lifetime tier for the builder community around the cosmos engine | Medium — lifetime pricing caps LTV; cap seat count |
+| D. OSS + paid cloud | Cosmos engine/MCP free to self-host; hosted+managed version paid | Medium — ops burden before product-market fit |
+
+**Recommendation: B now, A when W4 ships, C as a launch event, D last.** Templates monetize the existing audience without new infrastructure; the subscription needs the agent to exist first. Do not build billing before the thing billed for exists.
+
+## 5. Sequencing (each phase falsifiable before the next)
+
+1. **P1 — Substrate depth** (now → 100 cards): W1 continuous. Falsifier: no organic-traffic growth by 50+ cards → SEO thesis wrong, pause W5.
+2. **P2 — Agent surface**: W3 Cosmos MCP v0.1 + story/research presets. Falsifier: no external MCP installs/stars in 60 days → distribution problem, fix before W4.
+3. **P3 — Live agent + first revenue**: W4 site agent + monetization B→A.
+4. **P4 — Multi-surface**: W5 Expo app (content pipeline already proven), W2 3D layer.
+5. **P5 — Network layer**: W6 visualization; any chain/token work requires its own `/starlight-board` gate (sovereign-class).
+
+## 6. Guardrails carried forward
+
+- Facts: established figures only; estimates labeled; fast-moving numbers dated ("as of 2026") — Metrics Truth Rule applies to cosmos cards.
+- Myth and science both present, never mixed — the constellation pages set the pattern (three labeled layers).
+- Zero new runtime deps for content surfaces; three.js only at W2's 3D gate.
+- Every surface additive; no URL renames.
+- Board gates: anything touching SIP/attestation/chain economics is substrate-tier → `/starlight-board` before commit.
+
+---
+
+*Next session pick-up: W1 wave (planets + zodiac constellations) or W3 (Cosmos MCP v0.1 in starlight-cosmos-engine — consolidate the mcp-nasa-media / mcp-esa-webb / mcp-arxiv-space stubs into one `mcp-cosmos` server).*

site/content/cosmos/betelgeuse.md

@@ -0,0 +1,23 @@
+Betelgeuse is the red supergiant marking Orion's eastern shoulder — an M-type star so large that, placed where the Sun sits, its surface would extend past the asteroid belt and possibly out near Jupiter's orbit. It is one of the very few stars whose disk we can resolve from Earth, and almost everything about it comes with an honest error bar. That uncertainty is not a footnote here; it is the story.
+
+## A star with error bars
+
+Start with the distance: roughly 430 to 640 light-years, commonly quoted around 550. Betelgeuse is too far for clean geometric parallax and too bright and bloated for the usual tricks — its own convection cells shift its apparent position enough to contaminate the measurement. Because luminosity and physical size are derived from distance, the diameter estimates swing too: somewhere from around 640 to over 900 times the Sun's radius, depending on the adopted distance and the observing wavelength. A red supergiant has no crisp surface; its outer layers thin gradually into space, so "the edge" is partly a matter of definition. Mass estimates land around 15 to 20 solar masses. Any single confident number you see quoted for Betelgeuse is hiding a range.
+
+It is also a semiregular variable — it brightens and dims on overlapping timescales, roughly 400 days and about six years, driven by pulsations and enormous convection cells. Variability is its normal state.
+
+## How you measure a star that big and blurry
+
+In 1920, Albert Michelson and Francis Pease mounted a 20-foot interferometer on the 100-inch telescope at Mount Wilson and measured Betelgeuse's angular diameter — about 0.047 arcseconds. It was the first star other than the Sun whose size was directly measured. The trick was interference: combine light from two separated apertures, and the fringe pattern encodes spatial detail far finer than either aperture could resolve alone. Modern instruments like ESO's VLTI and ALMA use the same principle at higher precision, and they do not just measure the disk — they image it, revealing bright hotspots and an asymmetric, boiling surface.
+
+## The Great Dimming
+
+Between late 2019 and early 2020, Betelgeuse faded to roughly a third of its usual brightness — the faintest it had been in over a century of photometric records. Public speculation jumped straight to supernova. The instruments told a calmer story: Hubble caught a burst of hot material moving outward through the atmosphere in the months before the dimming, and the VLT's SPHERE imager showed the star's southern hemisphere darkened behind an obscuring cloud. The star had ejected a mass of gas from its surface; as it cooled, it condensed into dust and blocked part of the disk from our line of sight. Betelgeuse recovered. The event was a rare, close-up look at how red supergiants shed mass — the process that seeds galaxies with heavy elements.
+
+## The supernova, honestly
+
+Betelgeuse will explode as a core-collapse supernova — but "soon" means astronomical soon, likely within the next 100,000 years or so, not next Tuesday. Nothing observed, including the Great Dimming, indicates an imminent explosion. When it does go, it will be spectacular and safe: at hundreds of light-years, it will briefly rival the Moon in brightness and be visible in daylight for weeks, with no meaningful hazard to Earth.
+
+## Why it matters to a builder
+
+Betelgeuse is a masterclass in working with uncertainty instead of hiding it. Its distance, size, and fate are all ranges, and astronomers publish the ranges — conclusions carry their confidence intervals with them, the way good benchmarks and estimates should. The 1920 interferometry result is the deeper lesson: when no single instrument can resolve the problem, combine small ones and exploit correlation — the same architecture as phased arrays, aperture synthesis, and every distributed system that beats a monolith. And the Great Dimming is an incident postmortem done right: competing hypotheses, discriminating observations, root cause published. Panic said supernova; instrumentation said dust.

site/content/cosmos/cassiopeia.md

@@ -0,0 +1,23 @@
+Cassiopeia is the bright W of five stars circling the north celestial pole opposite the Big Dipper — and it happens to contain the loudest object in the radio sky. Point a radio telescope anywhere beyond the solar system and nothing outshines Cassiopeia A, the expanding wreckage of a star that died about 340 years ago, as seen from Earth.
+
+## The W is a projection
+
+Like almost every constellation, the W is a line-of-sight accident. Caph (Beta Cassiopeiae) is an F-type star only about 55 light-years away. Ruchbah (Delta) sits at roughly 99 light-years. Schedar (Alpha), a K-type orange giant, is about 228 light-years out. Segin (Epsilon) is roughly 400 light-years away, and Gamma Cassiopeiae — the middle of the W, bright enough that it carries the astronaut-era nickname "Navi" — is an unstable, rapidly spinning B-type star at roughly 550 light-years that flings material into a disk around itself and varies in brightness unpredictably. A factor of ten in distance across five stars; the W exists only from here.
+
+## Two dead stars that changed the field
+
+In November 1572 a new star blazed in Cassiopeia, briefly outshining Venus. Tycho Brahe measured it carefully, showed it had no parallax — meaning it lay far beyond the Moon — and published the result. That single observation cracked the Aristotelian doctrine that the heavens were perfect and unchanging. We now know it as SN 1572, a white-dwarf (Type Ia) supernova, and its remnant is still expanding.
+
+Cassiopeia A is the younger sibling: a massive-star supernova whose light reached Earth around the 1690s, oddly with no confirmed contemporary sighting — the explosion was likely dimmed by surrounding dust. Its remnant, about 11,000 light-years away, is the brightest radio source in the sky after the Sun and a standard calibration target for radio astronomy. JWST images of Cas A now resolve the debris field in enough detail to map individual knots of newly forged oxygen, argon, and neon being thrown into the galaxy — supernova nucleosynthesis, photographed.
+
+## The queen on the throne
+
+In Greek myth Cassiopeia was the queen of Aethiopia who boasted that she (or her daughter Andromeda) outshone the sea nymphs. Poseidon's punishment structured a whole region of the sky: Andromeda chained for the sea monster Cetus, Perseus arriving to intervene — all neighboring constellations. Cassiopeia's own sentence was to circle the pole forever strapped to her throne, hanging upside down for half of every rotation. Arabic tradition read the same stars differently — as a kneeling camel, or as "the tinted hand," a henna-stained hand — and Chinese astronomy assigned them to Wangliang, a famed charioteer, with a nearby star as his whip.
+
+## The counterweight to the Dipper
+
+From mid-northern latitudes Cassiopeia is circumpolar, and it sits across Polaris from the Big Dipper — when one rides high, the other hangs low. That makes it the standard fallback for finding north: on autumn evenings when the Dipper scrapes the horizon or drops behind trees, the W is near the zenith, with its open side facing toward Polaris. Navigators also read the pair as a rough clock, since the two asterisms sweep around the pole once per sidereal day.
+
+## Why it matters to a builder
+
+Cassiopeia rewards redundancy and calibration thinking. Navigation by the Dipper alone fails for part of every night; adding a second landmark on the opposite side of the pole gives the system full coverage — the same reason you run failover across zones rather than scaling one. And Cas A shows what a good reference signal is worth: radio astronomers measure instruments against it precisely because it is bright, stable in position, and thoroughly characterized. Every system needs its Cas A — a known-good source you trust enough to calibrate everything else against.

site/content/cosmos/crux.md

@@ -0,0 +1,21 @@
+Crux, the Southern Cross, is the smallest of the 88 constellations and the most heavily used piece of sky in the southern hemisphere. Four bright stars in a compact kite shape do the job that Polaris does in the north — except there is no bright south pole star, so Crux earns its keep as a pointer instead of a beacon.
+
+## Four stars, four different distances
+
+The cross is another line-of-sight pattern, not a physical group. Acrux, the foot, is a multiple system of hot B-type stars roughly 320 light-years away. Mimosa (Beta Crucis), the left arm, is a B-type giant at about 280 light-years. Gacrux, the top, breaks the pattern visually and physically: it is an M-type red giant only about 89 light-years away — the nearest red giant to the Sun — and its warm color is obvious to the naked eye against its blue-white neighbors. Delta Crucis rounds out the cross at roughly 345 light-years. Four stars, spanning a factor of about four in distance, unrelated except from our vantage point.
+
+## The Coalsack and the Jewel Box
+
+Crux sits in one of the richest bands of the southern Milky Way, which makes its two flagship deep-sky objects a study in contrast. The Coalsack is a dark nebula roughly 600 light-years away — a cloud of dust dense enough to blot out the starfield behind it, visible to the naked eye as a black hole punched in the Milky Way's glow. Just off Mimosa lies the Jewel Box (NGC 4755), an open cluster about 6,400 light-years away, only some 14 million years old, where a single orange supergiant sits among dozens of hot blue stars — one of the finest small-telescope targets in the sky.
+
+## The Emu, the flags, and a lost northern star
+
+Aboriginal Australian astronomy reads this region in a way European tradition never did: the Emu in the Sky is a constellation made of darkness, not stars, and the Coalsack is its head, with the body stretching along the dust lanes toward Scorpius. Its seasonal orientation tracked emu breeding cycles — a calendar drawn in dark nebulae. European tradition claimed the stars late; Ptolemy catalogued them as part of Centaurus, and precession has since carried them below the horizon for most of Europe, so early modern navigators encountered the Cross as a discovery. Today it anchors national identity across the southern hemisphere: the flags of Australia, New Zealand, Brazil, Papua New Guinea, and Samoa all carry it.
+
+## Finding south
+
+The method is mechanical. Extend the long axis of the cross, from Gacrux through Acrux, about four and a half times its own length. That point is close to the south celestial pole; drop a vertical line from it to the horizon and you are facing south. Navigators cross-check with the nearby Pointers, Alpha and Beta Centauri, to avoid confusing Crux with the larger, fainter False Cross nearby — a mistake that has genuinely misled sailors.
+
+## Why it matters to a builder
+
+Crux is the pattern for deriving what you need from what you have. The southern sky offers no Polaris — no single marker sitting on the answer — so navigators built a reliable procedure that computes the pole from visible reference points, plus a known failure mode (the False Cross) and a cross-check to guard against it. That is robust system design: when no ground truth is directly observable, derive it, and ship the validation step with the procedure. The Emu adds a second lesson — the information was in the dark regions all along, and an entire tradition of astronomy was built by attending to what the standard schema treated as background.