Decisions

Question

R2-DECK-LAYOUT DL1 playtest (dev build v0.4.0-dev.34-home-tiers-60-g45c0c91d, PR #287, frames refs/deck_layout_bow/dl1/): accept the layout A blockout as built (Voyage and Crew desks 0.6 m bowward, chair keep-out 0.5 m) and carry on to DL2?

Ruling

ACCEPTED (Mark, attended 2026-10-20): "very nice". Direction for DL2 onward: (1) with fixed viewpoints, author eclipse and transition framings (a body passing the star, the starbow forming, arrivals) as set pieces; (2) light sources carry the atmosphere: the real star lights the bridge when one is close (R1-SHIP-STAR-LIGHT), and when nothing close by is available the spire is the bridge's own in-world light source, replacing the abstract ship-fixed broad key.

Provenance

PR #287 head cd2d0d74: rig/full green; deck-layout-eval, camera-test 187, camera-forward-test 22, make golden 393 ok.

stapledon-mission.md, line 210 PR #287

Question

R2-CAMERA-FORWARD F1 playtest (dev build v0.4.0-dev.34-home-tiers-38-g0a787245, PR #282, review frames refs/camera_forward/f1/): accept the Bow view as built (extra occluder fades, captain ~81 px) and build ship assets against the Bow and Astern views?

Ruling

Accepted (Mark, attended 2026-10-20): "this looks great. we can do more ship building assets now assuming these". F1 Bow/Astern views are the basis for ship assets; finish F2-F3; re-plan R2-DECK-LAYOUT (D-90) and the Blender set dressing for the Bow view (spawn facing the lift goes to the deck plan).

Provenance

PR #282 (feat/camera-forward 0a787245): make test, camera tests and make golden green on the Studio; 8 review frames + walk contact sheet at gs://stapledons-voyage-assets/refs/camera_forward/f1/.

stapledon-mission.md, line 208 PR #282

Context

R2-CAMERA-FORWARD (game PR #281) measured that with the forward pole straight overhead (D-14, "up = forward"), no camera can hold the deck and the journey in one frame. At the fixed 78° field of view the pole is in frame only at a pitch of 51° or more, and from there no deck is visible. Interior gravity is the generator's comfort 1 g (D-11, physics/higgs-bubble.md HB-6), not thrust, so nothing in the physics ties "up" to the direction of travel.

Ruling

In Mark's words

shall we chaneg it so the direction of travel is the bubble moving sideways? > just for asethetic reasons? may make it all simpler? He then chose "Into the screen".

Rationale

Pillar 3 (Time Has Emotional Weight) and Pillar 5 (Grounded Strangeness): the journey ahead is on screen while the player walks, not one key away. Pillar 4 (The Ship Is Home): every open level's bow edge sees the forward sky, and its stern edge sees the sky go dark. The physics lesson now lives on every deck, not only at the bridge.

• Records ledger: D-92: The ship travels along the deck, into the screen

design-decisions.md, line 1530 (pinned 42a99a7) design PR #282 design PR #286 design PR #281

Question

Approve the R2-CAMERA-FORWARD plan (PR #281: design_docs/planned/r2/camera-forward.md + -sprint.md, sprint JSON awaiting_approval; design repo PR #30) and its open questions OQ1-OQ8.

Ruling

APPROVED (Mark, attended 2026-10-19): "yep looks great approved", with every recommendation: OQ1 Bow view candidate A (pitch -8 deg, 9 m back, 5.5 m up); OQ2 planet facing is the stop turn that brings the body into the Bow frame, a Target key later for free navigation; OQ3 Astern on key 9, no side view; OQ4 zoom levels Z1-Z3 removed for players, kept as developer presets, revisited after the dev build; OQ5 fade the spire, never turn the camera; OQ6 the Commons stays legacy; OQ7 inset Astern window later; OQ8 the painted-plate developer interior stays legacy this sprint. First pause point: a playable dev build after F1.

Provenance

Attended session 2026-10-19: bow candidates measured (gs://stapledons-voyage-assets/refs/camera_forward/bow_candidates/); independent Sonnet plan reviews 71, 76, 79 PASS.

stapledon-mission.md, line 207 PR #281 PR #30

Question

Ship orientation (amends D-14 "up = direction of travel throughout"): with D-91 fixed views, the forward pole straight overhead cannot share a frame with the deck (R2-CAMERA-FORWARD, PR #281, measured). Interior gravity is the generator's comfort 1 g, not thrust (D-11, physics/higgs-bubble.md HB-6), so travel may run along the deck. Change it, and which way does the default view see the travel?

Ruling

Provenance

Attended session 2026-10-19, PR #281 design review.

• Also recorded as: Travel Runs Along the Deck, Into the Screen

stapledon-mission.md, line 206 PR #281

Question

Camera model after the dev.33/dev.34 playtest (amends D-83 and R2-CAMERA, design_docs/implemented/r2/camera-fixed-views.md): the shipped camera swings to an authored yaw per sector (16 cells around the spire) as the captain walks, so the ship turns under the player, movement keys change screen direction, and the forward sky and the destination leave the frame. What should a fixed camera do as the captain walks?

Ruling

Provenance

Attended session 2026-10-19, Mark playing dev.34; R2-DECK-LAYOUT (D-90) paused at DL1 on feat/deck-layout (7863c831).

stapledon-mission.md, line 205

Question

Approve the R2-DECK-LAYOUT plan (PR #279: design_docs/planned/r2/deck-layout.md + -sprint.md, sprint JSON awaiting_approval) and its open questions OQ1-OQ8 (who models, style reference, captain chair as start and focal point, keep the spire needle, bridge and Commons scope, baked spire light, two layout candidates, Blender bpy scripts as authoring tools outside the game Python policy).

Ruling

APPROVED (Mark, attended 2026-10-19): "great please proceed - approved", with every recommendation: OQ1 an agent models through Blender bpy scripts, Mark may tweak; OQ2 the sprites' illustrated finish; OQ3 the captain chair is the start and focal point; OQ4 keep the spire needle and move props (fixed forward-view position as fallback); OQ5 bridge and Commons only; OQ6 bake the spire light; OQ7 two bridge layout candidates; OQ8 the Blender scripts live in ~/dev/blender as authoring tools outside the game's Python policy. Review points stay: Mark picks the layout after DL1/DL2 and approves the DL3 style frame.

Provenance

Attended session 2026-10-19: before contact sheet gs://stapledons-voyage-assets/refs/deck_layout/before/; pillar net +5; independent Sonnet plan review 74 then 86 PASS, 0 blocking.

stapledon-mission.md, line 204 PR #279

Question

Approve the R2-SIMILAR-S0 sprint plan (row L9, semantic-memory S0 similar worlds; PR #272: design_docs/planned/r2/semantic-memory-s0.md + -sprint.md, sprint JSON awaiting_approval). It deviates from L9 in two places: a top-level similar section in the new_game reply rather than system.similar (system membership changes with ship position), and emission from sim/protocol.ail with sim/core.ail untouched (protocol sessions already run strict). Which do you want? A — approve the plan as written; the loop executes S0.1–S0.3 (headless, no visible change) in later iterations. B — approve, but nest it as system.similar and/or re-export through sim/core.ail (say which); the loop amends the plan first. C — hold S0.

Ruling

DECIDED (Mark, attended 2026-10-19): A, approve the R2-SIMILAR-S0 plan as written (top-level similar section, emitted from sim/protocol.ail, sim/core.ail untouched); the loop executes S0.1-S0.3.

Provenance

Iteration 38 (loop, 2026-10-19): designer Opus wrote the doc (pillars net +7); quorum r1 BLOCKED (3 present), one revision, r2 BLOCKED 1 pass / 2 reject on missing verification rows, fixes applied under the narrow-refinement carve-out, gpt6-1-sol absent (unreachable) both rounds; plan by codex gpt-6.1-sol; independent Sonnet review 83/100 PASS, 0 blocking.

stapledon-mission.md, line 199 PR #272

Question

S1 render sign-off for M4.6rR3 (PR #265): the 3D-ship S1 set (7 states × deck and forward views, contact sheet) is public at https://storage.googleapis.com/stapledons-voyage-assets/refs/m4_ship_s1/contact_sheet.png. Physics gate 2 needs you to open it. Which do you want? A: signed off; the loop merges #265 after its CI is green (re-capturing first on the merged tree if attended #264 has landed, since #264 changes the scene the capture drives). The findings it shows go into the queue (flat white Earth and α Cen A discs, dust flashes at rest in the home frame, faded card ghosts, a hull spire in every forward view, cards over about 30 % of transit frames, no separate glow at 3.35e-5 W/m²). B: changes needed before sign-off (name which frames or views); the loop re-captures and asks again. AC12 (bench p99 < 16.7 ms) misses at HEAD in either case, and the parent plan (R3-8) puts it at P-land as row 7m.

Ruling

DECIDED (Mark, attended 2026-10-19): A, signed off. The loop merges #265 after its checks are green, re-capturing on the current tree first; the listed findings go into the queue as scene polish.

Provenance

Iteration 36 (loop, 2026-10-16): design design_docs/planned/r1/m4-6rr3-ship-renders-bench.md (pillar net +1); quorum PROCEED at 1 of 3 present; plan codex gpt-6.1-sol; executor codex gpt-6.1-sol (through an Agent-tool role), AC5 report design_docs/planned/r1/m4-6rr3-ship-renders-bench-report.md (15 frames opened); independent eval Sonnet 85 PASS, 0 blocking.

stapledon-mission.md, line 198 PR #265 PR #264

Question

M4 AC4 says both clocks and the ISM readout show on every transit sample. On the 3D ship, during braking, the Braking notice and the medium-boundary notice take both expanded card slots and fold the transit card (with its ISM readout) for about 2.8 s (42 of 716 samples). Which do you want? A — amend AC4: a folded transit card under those two named notices counts as shown (the ratchet already allows exactly this case, nothing else). B — keep AC4 literal: change the card priority so the transit card never folds in transit (a visible HUD change, attended).

Ruling

DECIDED (Mark, attended 2026-10-19): A. Amend AC4: a transit card folded under the Braking notice and the medium-boundary notice counts as shown; the ratchet keeps exactly that case pinned, nothing else.

Provenance

Iteration 29 (loop, 2026-10-14): make display-audit on the PR #232 scene run, display-audit known: transit-ISM 42, folding cards Cards/Card_medium_notice + Cards/Card_notice (Braking); entry in tests/fixtures/display_audit/known_findings.json; independent eval judged the exception tight (eval_R1-M4-JOURNEY_M4.5bR3_round_1.json).

stapledon-mission.md, line 191 PR #232

Question

R2-CAMERA zoom ruling (Mark, dev.31 playtest, 2026-10-16): "I think we need to turn off zoom in and out for players, and switch to fixed zoom levels instead. we may want to also start having more fixed camera angles and not be 1st person but 3rd person only." Free player zoom and mouse look, or fixed levels and authored third-person views?

Ruling

DECIDED (Mark, attended 2026-10-16, "yes continue" on the camera plan with all nine recommendations; amended 2026-10-17): free player zoom and mouse look are removed; three fixed zoom levels (Z1 Close, Z2 Deck default at rest, Z3 Whole level); third person only (console focus stays a framed screen); no player camera rotation in R2; automatic situation cues that any input ends, prompts under reduced motion; no exterior view this sprint; perspective 78 deg; a developer free camera kept (F8 / --free-camera). Amendment 2026-10-17: "Sky forward under way" (under way Sky forward, at a guided stop Sky side, at Sgr A* the hole view; any zoom key returns to Z1-Z3). Supersedes D-63 for players.

In Mark's words

yes continue

Provenance

Attended, 2026-10-16. Design design_docs/implemented/r2/camera-fixed-views.md; plan PR #261; implementation PR #270 (evaluation 79/100 PASS).

stapledon-mission.md, line 197 PR #261 PR #270

Context

The Archive becomes a crew actor in the crew simulation (stapledons-godot archive-npc A1; crew_core 0.6). [archive-crew-trust.md](../features/future/archive-crew-trust.md) scores trust on −100..100, while the crew simulation scores all trust on 0..100 with 50 neutral.

Ruling

design-decisions.md, line 1479 (pinned 42a99a7)

Question

Crew names and faces (portraits trial #238; refines D-74): generated international names don't know the face they're paired with (the trial's analyst face is a woman named Santiago Navarro). Should names fit the face, faces be drawn to fit the name from per-role pools, or each featured face keep one canonical name?

Ruling

DECIDED (Mark, attended 2026-10-14): the name fits the face. Each featured portrait records a name region and gender, and the D-74 generator draws a name that fits that face: the same faces every game, varied names. People beyond the featured cast (successors, the population) get a face generated to fit their drawn name (D-7 generate on first use).

Provenance

Attended, 2026-10-14. Seen in renders/crew_console from the trial; D-74 draws names in sim/crew_names.ail from 14 regions with no face input.

stapledon-mission.md, line 194 PR #238

Question

Crew portraits (portraits trial #238): the game's crew include a Counselor, but the generation-cast plan and the Codex art handoff (stapledons-design art/handoff-codex-art.md) have no counselor identity; the trial shows the Diplomat's face. Which identity is the Counselor?

Ruling

DECIDED (Mark, attended 2026-10-14): the Diplomat is the Counselor. The game's counselor role takes the Diplomat identity (departure 62, embarks with an adult child and grandchild); its emotion-portrait set comes with the five departure identities in the Codex handoff part (b) step 2; no new counselor face.

Provenance

Attended, 2026-10-14. Raised by the portraits trial (PR #238): data/ai/cast/counselor.json has no portrait set; the handoff's part (b) emotion sets are keyed by cast identity.

stapledon-mission.md, line 193 PR #238

Ruling

Rationale

Grounded Strangeness, because one named hand-wave replaces a silent one. Choices Are Final, because every departure burns part of the ship. Ship Is Home, because the household loop is real and fragile.

• Records ledger: D-69: Bubble resources and the drive's energy (stapledons-design#20, features/next/bub

design-decisions.md, line 1381 (pinned 42a99a7)

Question

Crew display names (R2-CREW C5; refines D-70 OQ2, which kept role ids and put display names in the cast files): role names ("The Pilot") or personal names, and where do personal names come from?

Ruling

DECIDED (Mark, attended 2026-10-13): 'yeah we will name them, generated from a big list of international name combinations'. The featured crew get personal names, generated per new game from a large list of international given-name and family-name combinations. Role ids stay the internal entity_id (D-70 OQ2); the cast files keep the role name as the fallback.

Provenance

Attended. Raised in the R2-CREW C5 status (2026-10-13).

stapledon-mission.md, line 188

Question

R2-CREW crew integration open questions (design_docs/planned/r2/crew-integration.md, PR #221): OQ1 crew turn length (ship-day or ship-week); OQ2 featured crew names; OQ3 first scenario aboard; OQ4 how the crew see the D-69 drive mass bill; OQ5 default captaincy and switching; OQ6 the ergosphere scenario heat imbalance (lab REVIEW).

Ruling

DECIDED (Mark, attended 2026-10-13): OQ1 one crew turn per ship-day; OQ2 role ids stay the entity_id, display names go in the cast files; OQ3 last-messages-home comes aboard first; OQ4 the drive mass bill is a ship event shown to the crew, readout only in R2; OQ5 consent by default, orders an explicit journalled choice at the Voyage console (adds a run-time policy command to crew_core CI-0); OQ6 the ergosphere scenario gets a third heat user before it comes aboard (lab authoring; not in the R2 slice). Folded into the design doc, merged as #221 (cd45aff).

Provenance

Attended. Raised in the crew integration design (PR #221, 2026-10-13).

stapledon-mission.md, line 184 PR #221

Question

Bubble resources and the drive's energy (stapledons-design#20, features/next/bubble-resources.md §4): the guided tour's interstellar legs cost ~3,100 kg of mass-energy, which no real fuel or radiator can supply, so the drive is already a second hand-wave. Options: A physics caps cruise near 0.9995c; B the generator turns interior mass into drive light; C total conversion for everything. Also: m_eff value, and may the wall's shape change?

Ruling

Provenance

Attended. Raised in the bubble-resources design (stapledons-design#20, 2026-10-13).

• Also recorded as: The Generator Is the Drive; One Closed Ledger of Resources

stapledon-mission.md, line 183 design PR #20

Question

In-world voice: may the Archive and the news from home say that something is invented, approximated or a game?

Ruling

Provenance

Attended.

stapledon-mission.md, line 181

Question

Approve sprint-plan revision round 3 of R1-M4-JOURNEY (design_docs/planned/r1/m4-first-journey-sprint.md § Revision round 3; D-52 A / D-56 / D-57): M4.3b closed by absorption into R1-SHIP-UI; M4.5's scene half on the default 3D ship (plain launch → Board the ship → consoles; walking charged in the ≤360 s proxy); M4.6's gated half as ship goldens G-M4-S2/S3/S4, 3D-ship S1 captures and bench SCENE=bridge (iso-only G-M4-1 dropped for the ship); two attended hooks H1 (record seam in demos/ship_geometry_demo.gd, ~4 LOC) and H2 (news + legacy panels at the Archive terminal in demos/ship_consoles.gd, ~50 LOC); S1 render sign-off before the renders PR merges. Options: approve (H1/H2 attended, then the loop builds M4.5aR3 and M4.6gR3) / approve with changes / reject (the loop works triage L1–L8, R1, R2).

Ruling

APPROVED (Mark, attended 2026-10-12): sprint-plan revision round 3 of R1-M4-JOURNEY as recommended. H1 (record seam in demos/ship_geometry_demo.gd) and H2 (news and legacy panels at the Archive terminal in demos/ship_consoles.gd) are done attended first; then the loop builds M4.5aR3 and M4.6gR3 on the default 3D ship and parks for Mark's S1 render sign-off before the renders PR merges.

Provenance

Iteration 24 (2026-10-12): planner pi:ollama/kimi-k3:cloud; evaluations eval_R1-M4-JOURNEY_round3-plan_round_1.json (84, 1 blocking, fixed) and …round_2.json (90 PASS, 0 blocking).

stapledon-mission.md, line 180

Context

Mark found the guided outer-system arrivals and their subsequent inward body legs repetitive and asked for a shorter tour of interesting views. The position audit found real 1,000-AU inward journeys of about 5.8 light-days, rather than a coordinate error. The existing D-54 finite-star planner can reach the close view in the interstellar leg itself.

Ruling

Rationale

Grounded Strangeness keeps the views tied to real positions and the existing planner. Time Has Emotional Weight keeps the two clocks honest across the shorter itinerary. Choices Are Final keeps every travel leg a genuine checked commitment. Callisto remains both an interesting moon view and the clearance leg needed after the close Jupiter stop.

• Records ledger: D-65:

design-decisions.md, line 1339 (pinned 42a99a7) design PR #201

Question

D-61 follow-up: accept the R1-ISM-DUST P2 afterglow, sensitivity/default dust assumptions, per-medium glow and Archive drafts after the concrete dev-build and render review?

Ruling

ACCEPTED (Mark, attended 2026-10-10: 'Accept the current P2 choices'): retain the current afterglow, 0.5 m spot radius and 0.2 s decay, labelled a game approximation; accept the sensitivity/default dust assumptions (epsilon 1e-10, no disputed radar-meteor large-grain tail), per-medium glow, new lism-1 baselines and the Archive drafts. This completes the approved P2 checkpoint for PR A. The standing instruction is to merge into main after technical gates pass, then update gallery/news with recent features, including black-hole orbit and Saturn arrival videos. No performance or scientific criterion is waived; dense-cloud PR B remains deferred with failed comparisons.

In Mark's words

Accept the current P2 choices

Provenance

Attended 2026-10-10. Review build v0.4.0-dev.25-free-nav-64-gefadd459; named renders/HUD and replay hashes in data/ism/review-baselines.json; Archive drafts in stapledons-design PR #12.

stapledon-mission.md, line 176 design PR #12

Question

Approve the ISM structure and dust-impact plan (PR #190, R1-ISM-DUST), the speed-limit gameplay and the package publishes.

Ruling

Provenance

Attended.

stapledon-mission.md, line 175 PR #190

Question

Physically grounded bubble-wall effects at speed: real interstellar medium structure and dust-grain impacts.

Ruling

Provenance

Attended.

stapledon-mission.md, line 174

Question

Approve the ship UI plan (PR #181, R1-SHIP-UI) and decide how it fits the loop's M4.3b.

Ruling

Provenance

Attended.

stapledon-mission.md, line 171 PR #181 PR #171 PR #182

Question

Starmap single truth: which position wins when catalogues disagree (precedence rule truth-1, PR #145).

Ruling

DECIDED (Mark, attended 2026-10-07): 'yes trust cns5s I guess i dunno not really qualified' - accepted the recommendation, not an expert judgement. Rule truth-1: where CNS5 judged a star (including its 69 choices of Hipparcos or DR2 over a poor EDR3 solution, RUWE >= 1.68), its parallax stands; Gaia is used for stars CNS5 never matched. Navigation moves for one star (CNS5:138). Revisit if a better rule or Gaia DR4 changes the audit.

Provenance

Attended.

stapledon-mission.md, line 165 PR #145

Question

Star colour in Auto view, and the final approach to a giant star (dev.19 review).

Ruling

DECIDED (Mark, attended 2026-10-07): 'lets have auto mode dim enough to have pretty star colours, and yes lets make a longer approach for bigger stars - these are all good gameplay touches'. Auto view dims a star's disc to just above lit bodies and lifts its own blackbody chroma (labelled on the HUD); Realistic is untouched. A giant's final approach lasts 90 s from where it is 25 deg across.

Provenance

Attended.

stapledon-mission.md, line 163

Question

The bubble-wall glow at high in-system speed, and go-ahead for finite destination stars.

Ruling

DECIDED (Mark, attended 2026-10-07): keep the glow as modelled ('the glow is great - it helps show how it is'); no Auto-view fading of it. Go ahead with the finite destination stars now ('yes please fix the stars now').

Provenance

Attended quotes above.

stapledon-mission.md, line 162

Question

Which ship scene does R1's first journey (bar clause 4) run in? The approved M4 plan (D-22) builds Track B on the isometric --interior composite (D-6): M4.3b moves the journey HUD and transit into it, and M4.5's playthrough audit and M4.6's gated goldens and S1 renders run there. Since the unified painted ship (attended 2026-10-05) the default launch opens the 3D painted ship, which already shows live boost, cruise and braking, and --interior is a developer reference mode. Options: (A) the unified 3D ship: re-plan M4.3b, M4.5 and M4.6's gated half against it, and stop building journey work into --interior; (B) keep the plan: build M4.3b into --interior and run M4.5 and S1 there, so clause 4 is proven in a scene players don't see by default; (C) split: M4.5's audit runs in whatever the default launch opens, and --interior keeps only its existing smoke test.

Ruling

DECIDED (Mark, attended 2026-10-08): A, the unified 3D painted ship. Re-plan M4.3b, M4.5 and M4.6's gated half against the default 3D ship; stop building journey work into --interior, which stays a developer reference mode with its existing smoke test.

Provenance

Loop iteration 20 (2026-10-07), Gate 2 reality check: design_docs/planned/r1/m4-unified-painted-ship.md § Approved continuation (one normal entry; --interior kept as a developer reference mode); the M4.3b, M4.5 and M4.6 rows in .ailang/state/sprints/sprint_R1-M4-JOURNEY.json; D-6.

stapledon-mission.md, line 166

Question

Ship geometry demo scope and camera consistency.

Ruling

APPROVED (Mark, attended 2026-10-04): bridge overlook and one-tier lift demo, with one meaningful perspective shared by stars, existing SR, play area and background geometry; future GR must use that same viewpoint.

Provenance

Approved design/sprint: design_docs/planned/r1/m4-ship-geometry-demo.md; isolated demo/ship-geometry-seven implementation branch.

stapledon-mission.md, line 150

Question

Boundary clearance in the selected seven-layer geometry reference.

Ruling

DECIDED (Mark, attended 2026-10-04): retain the proposed 5 m minimum clearance.

Provenance

Attended ship-geometry review; art/ship-dimensions-v2/layout.json and website/docs/ship-layer-reference.mdx.

stapledon-mission.md, line 149

Question

Bubble-ship major tier count for the geometry/art study.

Ruling

DECIDED (Mark, attended 2026-10-04): seven major layers are chosen; remove the twenty-deck option from the review; architecture within those layers will include buildings, forests and other spaces.

Provenance

Attended ship-geometry review; art/ship-dimensions-v2 and website/docs/ship-layer-reference.mdx.

stapledon-mission.md, line 148

Question

Godot 4.7 cannot play Ogg Opus; keep a WAV copy beside each voice blob?

Ruling

Mark's rulings (attended 2026-10-03): keep the WAV playback copy, the Ogg canonical, with the WAV verified to re-encode to the Ogg (make ai-playback-verify, in make test).

Provenance

Attended ruling during the AI.10a Medic style frame, recorded in the sprint plan.

• Refines: D-7 (not shown here)

ai-service-foundation-sprint.md, line 897 PR #91

Question

When should the portrait swap relative to the voice segment, and how long is the crossfade?

Ruling

Mark's rulings (attended 2026-10-03): swap timing 150 ms early, so swap_offset_ms defaults to -150 with the 120 ms crossfade (tests and capture updated; the rehearsal log does not carry the offset and is unchanged).

Provenance

Attended ruling during the AI.10a Medic style frame, recorded in the sprint plan.

• Refines: D-7 (not shown here)

ai-service-foundation-sprint.md, line 894 PR #91

Question

Which of the three captain silhouette proposals becomes the captain?

Ruling

Mark picked proposal A in the attended task on 2026-10-03. The eight final views preserve that identity, costume language and cream/ochre/violet palette. Final set status is proposed for visual review; the identity choice is picked. AI-generated (gpt-6.1-sol); no copyright claimed (D-33).

Provenance

Recorded in the captain set's generation notes (game PR #98); the eight final views are proposed for visual review, the identity choice is picked.

• Refines: Art Handoff Follow-ups: Medic v2 Stays; Captain Stages; Art Channel

generation_notes.md, line 3 PR #98

Context

The M4.2 evaluation found the bubble-wall glow pole at 0.99c 130× brighter than the 23.5 mag/arcsec² dark sky, using the canon ε = 10⁻⁹ and a placeholder equal-energy white spectrum (182.6 lm/W). At that level it drowns the starfield, which contradicts "faint by design". physics/higgs-bubble.md gave no spectrum.

Ruling

Rationale

- Pillar 5 (Grounded Strangeness): the colour follows the physics of the impacts, red at the default cruise and blue-white at the top. Only one admitted hand-wave remains, the bubble. - Pillar 3 (Time Has Emotional Weight): the starbow, the signature view of relativistic travel, stays readable at every speed. - The glow is a motion cue from about 0.997c up. It appears as the player pushes into the costly end of the slider, and fades as the ship brakes.

design-decisions.md, line 1203 (pinned 42a99a7)

Context

Mark approved bridge v1 (game PR #97) for the playable demo, and asked for a final bridge as good as the captain avatar sprites (game PR #98). The Blender thread's quality study (game issue #93) compared: - (a) a textured toon GLB; - (b) (a) plus a richer kit; - (c) an illustrated plate projected through the fixed iso camera onto the play GLB, with a Godot capture of (c).

Ruling

Rationale

Pillar 4 (The Ship Is Home): the bridge has to look as hand-made as its people. Pillar 5 (Grounded Strangeness): only the spire's fixed light is baked, so everything that depends on speed stays physically live.

design-decisions.md, line 1110 (pinned 42a99a7) design PR #97 design PR #98 issue #93 design PR #9 design PR #10

Context

The Blender thread built bridge v2 with a look-down panorama showing the lower levels (ruling 4 above). That framing hid the starbow at 0.99c, and only a rim of the level below showed above the bridge's far rail.

Ruling

1. Keep the starbow. The bridge panorama prioritises the forward sky: it keeps v1's up-looking camera, painted with the v2 look. This supersedes ruling 4 of the bridge v2 entry above. 2. The bridge was never meant to show the lower levels, and its perspective can't. The sphere-scaled levels (ruling 6 above) are guidance for future areas, as later generations move down the ship: - each lower level is wider down to the equator, then narrower again; - each lower level is roofed by the underside of the level above. 3. Keep the moody spire-only lighting as built. 4. Ship both facing hints. 5. The palette shift is accepted.

design-decisions.md, line 1175 (pinned 42a99a7)

Context

Two new art threads start: a Blender thread (a Claude Opus agent) and a 2D art thread (Codex "Sol 6.1" agents, which have no Claude tooling). A read-only audit found bridge build-out v1 unmerged in the Blender repo, and conflicting character specs (2048 vs 1254 px, the age grid, filename patterns, Medic v2 vs cast_marker_v1).

Ruling

• Refined by: Art Handoff Follow-ups: Medic v2 Stays; Captain Stages; Art Channel
• Supersedes: the 2048 px portrait target in the character brief and the cast plan (now the generator's native ~1254 px, never upscaled); the fixed +0/+25/+50/+75 age grid (now stages 0/20/40/60 years since departure)

design-decisions.md, line 1006 (pinned 42a99a7) issue #1

Context

Mark answered the open questions on the 2026-10-03 art handoffs (design-repo PR #4).

Ruling

• Refines: D-16: Bridge v1 approved for build-out; Medic review stage closed; Art Handoffs for M4.2: Finish Bridge v1; Captain Sprite and Cast Portraits
• Refined by: The captain: proposal A, the Longline Navigator

design-decisions.md, line 1078 (pinned 42a99a7) design PR #4 issue #93 issue #1

Question

Licensing of our own AI-generated work: the captain and bridge art and generation notes say 'owned by Sunholo, Apache-2.0', but purely AI-generated output may not be copyrightable. Scope?

Ruling

DECIDED (Mark, attended 2026-10-03): art only, no claim: remove ownership/Apache claims from AI-generated art (captain, bridge, concept art) and label it AI-generated, no copyright claimed; the code keeps Apache-2.0; third-party CC BY credits stay

Provenance

assets/characters/captain/*, refs/bridge_v2 BRIEF, website/docs/credits.md, README.

stapledon-mission.md, line 147

Question

Forward glow spectrum: higgs-bubble.md does not define the colour of the wall glow (placeholder equal-energy white, 182.6 lm/W). Options: hot-plasma blackbody from the impact energy / neutral white / hydrogen recombination lines.

Ruling

DECIDED (Mark, attended 2026-10-03): the glow spectrum is a hot-plasma blackbody whose temperature is derived from the impact energy, computed in sunholo/relativity first (package-first), replacing the equal-energy white placeholder.

Provenance

M4.2 PR #102; interior/forward_glow.gd.

stapledon-mission.md, line 144 PR #102

Question

Forward glow strength (M4.2 evaluation): at the canon ε = 1e-9 the bubble-wall glow pole is 130× the 23.5 mag/arcsec² dark sky at 0.99c and drowns the starfield, contradicting "faint by design". Options: lower ε (~1e-11) / keep and fix the text / tie to the glazing follow-up.

Ruling

DECIDED (Mark, attended 2026-10-03): lower the canon wall conversion efficiency to about 1e-11 so the forward glow is faint and the starbow stays visible (design-repo canon change: higgs-bubble.md eps and dependent HB rows, the package check values and the sim).

Provenance

Merge approval (Mark, attended 2026-10-05): "can we merge in all the PRs and branches now?" authorizes landing the current PR #113 proposal, ε = 1e-10 with its checked spectral implementation and controlled replay re-freeze, together with the linked canon update. eval_R1-M4-JOURNEY-M4.2_round_1.json; test_physics.gd:885; m4-first-journey.md:362.

stapledon-mission.md, line 143 PR #113

Question

M1 sky mirror (M4.2 evaluation, confirmed four ways): Starfield.galactic_to_world is a reflection (det −1), so the whole M1 sky (stars, panorama, sky model, CMB) renders as a mirror image of the real sky; GPU-vs-CPU goldens could not see it because both sides share the map. M4.2 corrects it locally. Fix at the source now or later?

Ruling

DECIDED (Mark, attended 2026-10-03): fix the mirror at the source now in a dedicated PR: one right-handed rotation map used everywhere (the seven copies removed), the interior's local flip dropped, and every reference render regenerated and reviewed.

Provenance

eval_R1-M4-JOURNEY-M4.2_round_1.json; sky/starfield.gd:31.

stapledon-mission.md, line 142

Question

Forward CMB glare (M1.8 evaluation P1): at γ 707 the forward disc is ~2e8 cd/m² (an eighth of the solar disc) and the model adapts the eye and draws a clean dot with no glare, saturation or afterimage. Options: glare/saturation plus auto-dimming labelled glazing / HUD warning only / leave it.

Ruling

DECIDED (Mark, attended 2026-10-03): physically motivated glare/saturation around the forward disc plus auto-dimming labelled glazing, as a follow-up milestone (queue row 6b); M1.8 merges as is. R-e approved.

Provenance

eval_R1-M1-SKY-2-M1.8_round_1.json; sky/sky_meter.gd, sky/exposure.gd.

• Refines: D-25: Exposure anchors: Milky Way 20.8, deep space 23.5 mag/arcsec²

stapledon-mission.md, line 141

Question

Planet renders (Saturn, solar-system flybys, as in the retired Go build) are on no roadmap; the design repo only has Go-era feature docs (future/planet-rendering-polish, planetary-rings, planet-state-transitions, ailang-planet-ring-moon-data). Options: a new R1 milestone M5 / fold into M4 arrival / R2.

Ruling

DECIDED (Mark, attended 2026-10-03): a new R1 milestone M5 'Planets and flybys' (queue row 6c): AILANG-driven solar-system data, Godot planet rendering with rings and relativistic flyby views, Sol's planets plus the α Cen arrival scene; design doc first.

Provenance

stapledons-design/features/future/*planet*; legacy/go-engine-features/ailang-solar-system-demo.md.

stapledon-mission.md, line 140

Question

M1.5a exposure anchors flagged by the M1.5a evaluation: the Milky Way brightness anchor for the un-stretched NOIRLab panorama (executor used an uncited 20 mag/arcsec²), and the dark-sky surface brightness (22 mag/arcsec² is a ground-site value with airglow and zodiacal light; interstellar space is about 23.5–24).

Ruling

DECIDED (Mark, attended 2026-10-03): Milky Way anchor 20.8 mag/arcsec² cited to Leinert et al. 1998 (A&AS 127, 1, Table 16) and Duriscoe 2013 (PASP 125, 1370), replacing the uncited 20, with a Gaia/Tycho integrated-flux calibration as a later M1 item; dark sky moves to the deep-space value (~23.5 mag/arcsec², integrated starlight + diffuse galactic light + EBL, no airglow or zodiacal light) in M1.5a itself, with the AC8 naked-eye window moving accordingly (~6.9); R-d approved on the look (eye default, camera labelled).

Provenance

eval_R1-M1-SKY-2-M1.5a_round_1.json; sky/background.gd MILKY_WAY_MAG; sky/exposure.gd DARK_SKY_MAG.

• Refines: D-19: Remaining sky plan; the naked eye is the default exposure
• Refined by: D-27: Forward CMB glare and auto-dimming glazing

stapledon-mission.md, line 139

Question

α Cen distance in M4 (design OQ4, plan Q1): the HB rows and the Archive lore use 4.37 ly, but the catalogue after D-19 gives 4.32 ly, so the HUD and the Archive would differ by about 1 %. Options: a design-repo change setting HB-4 to the catalogue value and regenerating the dependent HB rows and entries from the package / keep 4.37 for the scripted checks and list the gap.

Ruling

DECIDED (Mark, attended 2026-10-03): change the design canon to the catalogue distance, 4.32 ly (HB-4 and the dependent HB rows and Archive entries regenerated from the package), so HUD and Archive agree.

Provenance

design_docs/planned/r1/m4-first-journey.md Open questions 4; plan Q1.

• Refines: D-19: Remaining sky plan; the naked eye is the default exposure

stapledon-mission.md, line 137

Question

Which generated voice speaks for the Medic?

Ruling

Voice: Aoede after an audition (three voices) (lands in AI.10b, ⏸ C).

Provenance

D-20 ruling 9 (Mark, attended 2026-10-02), as applied in the AI sprint plan; the audition is pause point C.

• Refines: D-20: Who pays for AI, and the player's spend ceiling

ai-service-foundation-sprint.md, line 81

Question

How are emotion markers written in generated text, and how is speech generated?

Ruling

Marker syntax: {emotion} (lands in AI.1). TTS granularity: Per segment, concatenated (lands in AI.10a).

Provenance

D-20 rulings 6 and 7 (Mark, attended 2026-10-02), as applied in the AI sprint plan.

ai-service-foundation-sprint.md, line 78

Question

AI text constraints flagged by the AI.2 evaluation: should no-digits apply to news and Archive text too, and what are the per-purpose length caps?

Ruling

DECIDED (Mark, attended 2026-10-02): no-digits applies to EVERY generated text (crew lines, news, Archive), so numbers on screen come only from sim fields and prose refers to them in words; per-purpose length caps are 280 characters for crew lines, 280 for news items (raised from 200), and 600 for Archive answers.

Provenance

eval_R1-AI-FOUNDATION-AI.2_round_1.json; sim/ai.ail maxCharsFor.

stapledon-mission.md, line 135

Question

M1 remaining-work plan (sprint R1-M1-SKY-2) and its open questions: alpha Cen A/B distance, default exposure, PR #43 merge order, and the planner's defaults (Hipparcos photometry fill, Python fallback retired, bright tier committed, map content, CMB in M1, D-6 canon wording).

Ruling

Provenance

design_docs/planned/r1/m1-remaining-sprint.md, Questions for Mark.

• Refined by: D-23: α Centauri at the catalogue distance, 4.32 ly; D-25: Exposure anchors: Milky Way 20.8, deep space 23.5 mag/arcsec²

stapledon-mission.md, line 133 PR #43

Question

M2 checkpoint R1 review (galaxy map and plan panel, v0.2.0-m2-map): which α Cen distance the map shows (catalogue 4.357 ly from 0.01-ly rounding vs check row 4.37 ly), star names, ratifying the executor-chosen scenario-parameter ranges, and whether to continue to M2.6b.

Ruling

R1 ACCEPTED, CONTINUE TO M2.6b (Mark, attended 2026-10-02): the map always shows the catalogue's own distance; full-precision positions come with the M1.2 catalogue tiers (alpha Cen A ~4.365 ly), and the 4.37 ly check row stays a package test, never shown in game; add common star names (alpha Centauri A, Sirius, Barnard's Star...) from a name table in M2.6b with the catalogue id as subtitle; the executor-chosen scenario-parameter ranges in the M2 design doc protocol section are RATIFIED as written.

Provenance

Review page https://claude.ai/artifact/WSyzsBATgJZVBJRSoa9E8E; release v0.2.0-m2-map; PR #28 (eval 89).

stapledon-mission.md, line 131 PR #28

Ruling

(1) The wall passes light of every energy; the crew is protected by real glazing and hull inside the bubble, which absorb the blueshifted UV and soft X-rays. No extra hand-wave. (2) The cruise slider reaches 0.999999c (γ ≈ 707); the ISM drag energy, which grows as γ·d, is the reason not to max it. The 10–20 default γ cap is retired; a cap remains a scenario parameter. (3) The boundary does not refract: light crosses undeviated, so the crew sees exactly the relativistic sky. (4) Bubble defaults: m_eff = 1 kg, boost proper acceleration 7.5×10⁵ g (about 5 ship-minutes to 0.999999c), inelastic glow fraction ε = 10⁻⁹; all scenario parameters.

• Records ledger: D-15: Glazing, the γ range, boundary optics, bubble defaults

design-decisions.md, line 972 (pinned 42a99a7)

Ruling

Bridge style frame v1 (2026-09-28, "ok its a good first pass") is approved for build-out. The Medic v2 images are superseded by the accepted illustrated cast. The exterior style frames remain to be reviewed. Art is modular data: the game runs on the current bundle (blockout, then bridge v1, then revisions), each delivery gets a review build, and tweaks arrive as data wherever possible.

In Mark's words

"I thought we did do art direction"; "looking to get something up so I can review it, and then we can tweak as we go? so we will have it modular enough to work with it?"

• Records ledger: D-16: Bridge v1 approved for build-out; Medic review stage closed

design-decisions.md, line 993 (pinned 42a99a7)

Question

Art status for M4: does bridge style frame v1 ('ok its a good first pass', 2026-09-28) count as sign-off for build-out (M4 stop S2), and the status of the exterior/Medic v2 and crew-exterior studies.

Ruling

DECIDED (Mark, attended 2026-10-01: 'I thought we did do art direction'): bridge style frame v1 is APPROVED for build-out, so M4 stop S2 is satisfied and a Blender agent may start ship-interior-blender-brief section 7 step 2; the Medic v2 images (style_revision_v2) are superseded by the accepted illustrated cast (cast_marker_v1); the exterior style frames remain to be reviewed.

In Mark's words

I thought we did do art direction

Provenance

~/dev/blender docs/stapledon/bridge_style_v1, style_revision_v2, crew_exterior_style_v1; art/README.md review state.

• Refined by: Art Handoff Follow-ups: Medic v2 Stays; Captain Stages; Art Channel
• Also recorded as: Art Status: Bridge v1 Approved; Iterate as Data

stapledon-mission.md, line 130

Question

Bubble follow-ups from the D-11 canon write-up: radiation shielding vs a wall passing light of every energy, the γ cap, boundary refraction, and M2 OQ6 bubble defaults (m_eff, boost_g, ε).

Ruling

DECIDED (Mark, attended 2026-10-01): (1) the wall passes light of every energy and the crew is protected by real glazing and hull inside the bubble, which absorb the blueshifted UV/X-rays; 'Radiation Shielding Automatic' (2025-12-06) is superseded in part, no extra hand-wave; (2) the cruise slider reaches 0.999999c (gamma ~707) with the ISM drag energy (~gamma*d) as the natural brake; the 10-20 default gamma cap is retired, a cap stays a scenario parameter; (3) the boundary does not refract, so the crew sees exactly the relativistic sky; (4) M2 OQ6 bubble defaults m_eff = 1 kg, boost proper acceleration 7.5e5 g, inelastic glow fraction epsilon = 1e-9, all scenario parameters.

Provenance

stapledons-design PR #2 open points; M2 OQ6.

• Supersedes: Radiation Shielding Automatic; Gamma Cap: 10-20 Default
• Also recorded as: Shielding, γ Range, Boundary Optics and Bubble Defaults

stapledon-mission.md, line 129 design PR #2

Question

M4 open questions (orientation, slice cruise speed, stand-off, calendar, news beat, art stop S2, playtest incl. bar clause 4, time while docked).

Ruling

M4 OQs DECIDED (Mark, attended 2026-10-01): no flip exists, up = direction of travel throughout; slice default cruise 0.99c for Sol -> alpha Cen (player can change it); journeys stop 1,000 AU from the target star; relative calendar; the news beat is a text panel at the Archive terminal; art stop S2 WAITS for Mark's approval of the Blender bridge style frame (no shipping on the spike blockout); R1 uses the scripted minimum-path proxy for the under-10-minutes bar and the 3-new-player human playtest moves to R2 (bar clause 4 amended); time while docked follows D-12 (no pause).

Provenance

design_docs/planned/r1/m4-first-journey.md Open questions.

• Also recorded as: Vertical Slice: 0.99c Default and a 1,000 AU Stand-off

stapledon-mission.md, line 128

Question

M2 remaining open questions (OQ3 calendar, OQ4 manual flight and pausing, OQ5 commit ritual).

Ruling

M2 OQ3-OQ5 DECIDED (Mark, attended 2026-10-01): calendar is relative ('Earth +6.003 yr') with epoch and start_age as scenario parameters, start_age 30; manual thrust only in diag sessions, in play the clock runs at rest at a fixed host rate with no pause while planning; commit ritual is one dialog showing both clocks and the years left at home with a 1.5 s hold.

Provenance

design_docs/planned/r1/m2-journey-core.md Open questions.

stapledon-mission.md, line 126

Question

M1.4a background source (M1 design open question 1): Option A on which panorama, NOIRLab noirlab2430b (E. Slawik, 10000×5000, CC BY 4.0) or ESO eso0932a (S. Brunier, CC BY 4.0)? Gaia EDR3 flux map ruled out (built from stars, no nebulae, CC BY-SA).

Ruling

OPTION A ON NOIRLAB (Mark, attended 2026-10-01: 'ok lets go with NOIRlab'): the background is the real NOIRLab noirlab2430b panorama (E. Slawik, CC BY 4.0, 10000x5000), with the point sources that are cross-matched to the shipped point layers (CNS5, GCNS, HIP2 V<7) masked and inpainted, and unmatched faint stars kept as pixels. ESO eso0932a and the Gaia EDR3 flux map are not used. The panorama is valid to about 50 ly from Sol (parallax); beyond that it is a stated post-R1 limit.

In Mark's words

ok lets go with NOIRlab

Provenance

Attended spike, branch m1.4a-background-spike, docs/m1.4a/README.md (matched-window crops, parallax validity note: a single panorama holds to about 50 ly from Sol).

stapledon-mission.md, line 124

Context

the log contradicted itself: isometric tiles (the original bridge design), first-person 3D (2025-12-18), and fixed AI-painted scenes (2025-12-20). The fixed-scene pivot happened only because AI images couldn't give consistent assets. Blender removes that constraint. Three feasibility spikes (stapledons-godot branch spike/iso-bridge, v1–v3) tested the options visually with the live physics.

Ruling

Rationale

Pillar 3 (Time Has Emotional Weight): the cosmos going dark around the lower decks at speed and blazing at the bow makes the journey visceral from inside daily life. Pillar 4 (The Ship Is Home): the player walks the ship among the crew. Pillar 5 (Grounded Strangeness): every star is where physics puts it, because the sky is rendered through the panorama's own camera.

design-decisions.md, line 700 (pinned 42a99a7)

Context

Mark reviewed the first Blender Medic character study and the first exterior style frame after accepting the bridge as a good first pass.

Ruling

In Mark's words

the character study is aweful. I think we will instead generate images that are much richer and expressive, and not use animation, just change expression as we go. so no blender for that or no? we can use mini avatars for them. the exterir looks more like a long egg rather than a perfect sphere the story prefers, and what is that tree that is outside the rest of the sphere? we can havea . forest level, makes sense. we also need some represtntaion of the bubble itself.

Rationale

The first character study did not deliver the emotional richness Mark wants from conversations. The exterior must communicate the established spherical ship and contain its inhabited forest visibly within that volume.

design-decisions.md, line 751 (pinned 42a99a7)

Question

Runtime AI operating model (ai-showcase open questions 1–3): keys and cost, providers, live voice.

Ruling

DECIDED (Mark, attended 2026-09-28). (1) KEYS AND COST: the PLAYER'S OWN KEY. The game ships PRE-GENERATED CORE CONTENT (the accepted founding cast, their portraits, avatars and voices, the Archive, and templated fallback lines), and live generation is OPT-IN with the player's key. Without a key the game is fully playable on the pre-generated core. (2) PROVIDERS: stay MODEL-NEUTRAL and swap models as needed through AILANG's provider routing; GEMINI is the default (text, image and TTS). (3) VOICE: NO LIVE real-time voice for now (no streaming conversation with the Archive; gemini_live deferred). Speech is PRE-RECORDED or generated per line and cached, and TEXT-ONLY mode is always available.

Provenance

Attended session 2026-09-28; features/ai-showcase.md §7.

• Also recorded as: Runtime AI Operating Model

stapledon-mission.md, line 122

Question

Ship-interior presentation for M4 (resolves the isometric / first-person / fixed-scene contradiction in the design docs: design-decisions 2025-12-18 vs features/scene-based-interior-navigation 2025-12-20).

Ruling

In Mark's words

yes this is more like it - lets go with these

Provenance

Attended session 2026-09-28; feasibility spikes v1–v3 on branch spike/iso-bridge.

• Supersedes: Pivot from Isometric to First-Person 3D; Interior Ship Experience: Scene-Based Navigation (recorded retroactively)

stapledon-mission.md, line 120

Question

Bright-star tier (M1 design open question 5): accept the proposed M1.2d, a Hipparcos (HIP2, VizieR I/311) tier for V < 7 stars not in GCNS or CNS5, with AC11 and teffFromBV in sunholo/relativity 0.3.0 (about +250 LOC)? Without it Rigel, Deneb and most naked-eye stars beyond 100 pc exist only as panorama pixels (option A) or not at all (option B).

Ruling

ACCEPTED (Mark, attended 2026-09-28: 'Yes I accept those stars'): add M1.2d, a Hipparcos (HIP2, VizieR I/311) tier for V < 7 stars not in GCNS or CNS5, with AC11 and teffFromBV in sunholo/relativity 0.3.0 (about +250 LOC), so Rigel, Deneb and the naked-eye sky beyond 100 pc are real catalogue stars.

In Mark's words

Yes I accept those stars

Provenance

Quorum round 1 (oc-glm-5-3), 2026-09-28; design doc Problem 5, rows V14 and M1.2d.

stapledon-mission.md, line 119

Question

White dwarfs in M1: approximate blackbody fit, flagged approximate (design open question 3).

Ruling

YES: approximate blackbody fit for white dwarfs in M1, flagged approximate in data and UI (Mark, attended 2026-09-28: "record decisions as ratified").

In Mark's words

record decisions as ratified

Provenance

Handover comment on #1; photometry 0.2.0 deliberately has no WD model.

stapledon-mission.md, line 118 issue #1

Ruling

The ship's visual style draws from French 70s science fiction comics, particularly Moebius (Jean Giraud), Philippe Druillet, and Métal Hurlant magazine. Key characteristics: (1) Organic/mechanical blend - technology that looks grown, (2) Saturated colors against vast emptiness, (3) Tiny humans against massive structures for scale contrast, (4) Clean flowing curves over sharp angles, (5) Cathedral-like spaces that evoke awe. This replaces the earlier 'clean functional sci-fi' aesthetic in bridge-interior.md.

design-decisions.md, line 633 (pinned 42a99a7)

Ruling

The ship is oriented vertically along the thrust axis. The spire runs from engines (bottom/aft) to observation deck (top/forward). Continuous 1g thrust provides artificial gravity - 'down' is toward the engines, 'up' is toward the direction of travel. Levels are stacked along this axis, with open sides facing outward toward the bubble boundary. This means when standing on any level, you look UP toward the bridge and DOWN toward engineering.

design-decisions.md, line 647 (pinned 42a99a7)

Ruling

Implement tiered visual physics based on ship speed: <0.1c uses boundary glow only, 0.1-0.3c adds faint aberration, 0.3-0.5c shows nearby star parallax, 0.5-0.9c has strong aberration (60°→26° cone), >0.9c produces extreme starbow effect. Dual view modes: 'Raw SR View' (physically accurate aberration/Doppler) and 'Navigation View' (computer-compensated with enhanced parallax). Local system parallax (planets, moons) is always visible when near objects - only stellar parallax requires high speeds.

design-decisions.md, line 600 (pinned 42a99a7)

Ruling

Levels are mostly open to the sides, providing direct views outward through the transparent bubble boundary to space. Players see: (1) Close-up visual effects with parallax at the bubble boundary, (2) Stars and planets beyond with SR/GR effects applied based on velocity, (3) The bubble glow from ISM impacts. This creates constant visual connection to the cosmos from daily life - you're always aware you're hurtling through space.

design-decisions.md, line 670 (pinned 42a99a7)

Ruling

The ship interior is 10-20+ levels radiating outward from the central Higgs Generator Spire. Each level serves different purposes: Command (bridge, observation), Residential (small houses, dwellings), Gardens/Cathedral (remembrance, culture), Industrial (engineering, fabrication), Commons (markets, gathering), Archive (AI shrine). Levels are mostly open to the sides with views outward. Navigation is via various lifts and ramps of different sizes arranged somewhat chaotically.

design-decisions.md, line 626 (pinned 42a99a7)

Ruling

Parallax visibility follows a natural lifecycle: DEPARTURE (rich local parallax from planets/moons, lasting minutes to days as you leave a system), INTERSTELLAR CRUISE (parallax desert - stars too far apart, motion cues from boundary glow and SR effects only), ARRIVAL (parallax returns as destination system objects become resolvable). This means 'artificial dust layers' are only needed during cruise phase - departure and arrival have natural 3D geometry.

design-decisions.md, line 619 (pinned 42a99a7)

Ruling

The observation deck / bridge at the top of the ship is the primary location for major decisions. Captain makes choices with cosmic backdrop - starfields, SR/GR effects visible.

design-decisions.md, line 495 (pinned 42a99a7)

Ruling

The "garden cathedral" in the outer shell is a gameplay-significant location. It's where: - Crew remember Earth and what it meant to live on a planet - Cultural rituals develop over generations - Philosophical conversations and processing happen - The bubble society's culture crystallizes Emotionally: "sad but happy" - bittersweet remembrance.

design-decisions.md, line 525 (pinned 42a99a7)

Context

Relativistic travel is central to the game, but players need to SEE it, not just read numbers. The question was how to make SR tangible without resorting to fake "warp streaks."

Ruling

Rationale

Serves Pillar 3 (Time Weight): Visual distortion makes time dilation visceral, not abstract. Serves Pillar 5 (Grounded Strangeness): Real physics that's actually visible. Serves Pillar 6 (Human Incompatibility): Disorienting visuals emphasize cosmic alienation—this isn't how humans evolved to see.

design-decisions.md, line 299 (pinned 42a99a7)