Dragonfly Titan GCS

October29 continuation: fan-to-duct transition. One generic open shell replaces the solid transition/7 mm flange gap. Estimated wall/throat clears retained rotors; other hardware and simulation unchanged. Part18 has a focused inspection detail.787 tests and desktop/phone layers pass; source, rejected trial, archive and samples saved. Published 700c882; Pages 40945342706 succeeded, 793 combined tests pass. Live part18 focus, estimate metadata,32 IDs and390px layout verified. Claude's incoming game files unchanged.

October29: part17 open fan housings. Six solid housing/flange shapes now have generic circular openings within the same outer envelopes, exposing existing rotors. No new fan identity, simulation or duct changes.771 tests and desktop/phone layer checks pass. Actual samples, source context, exact prior code and continuation notes saved. Owner clarified Claude owns Swarmfall; game/graph changes stay separate from this checklist. Published bd5e42b, Pages40940042344 succeeded; 782 combined tests pass. Live fan geometry,32 IDs/seven RF rows and390px no-overflow/errors verified.

Local October29 continuation: RF installation review. Seven closed evidence rows under Connection installation status distinguish RF, high-voltage power and actuator wiring. Existing parts/routes stay unchanged. Hidden gimbal feed remains unmodeled; the visible feed moves rigidly with the arm. Initial769 tests passed; RF is now included in publishedbd5e42b above.

Prior connection pass: Frontier radio bay lead. The existing branch now seats in a labeled EST shell/strain relief at the photo-labeled J22 face, with one fitted support. One estimated lead is refitted from14 to6 mm; its two network junctions and all other routes remain exact. Radio/RF geometry unchanged. Radio13 and Bay32 share the single link; RF-only focus omits it.761 tests and desktop/Thermal/390px checks pass. J22 circuit and actual flight connection are unconfirmed; endpoint status is geometric contact only. Published 89b4899; Pages 40815721341 succeeded, 766 combined tests pass. Live numbering/disclosures and no overflow/errors verified; public receipt saved.

October 28 correction: Compact blower tubing. Owner rejected the loose-looking loops in the first blower installation. Four shorter rigid-tube examples now use straight sections/local bends, two retention clips and short open exhaust spouts. Intake lengths shrink37–52%, exhausts61%. Unit exteriors, poses, selector and all upstream sample plumbing stay exact. Routes/supports remain estimated; no flight installation or pressure-loss claim. Before/after sample and prior source/archive retained. 751 merged tests pass; desktop/Thermal/390px phone have no overflow/errors. Published 6de0fab; Pages 40740446629 succeeded. Live compact-pipe disclosure, 32 numbered list entries and33 selector options verified; public sample/receipt saved.

October 28: Supplier-referenced DrACO blower pair. Real engineering-model photos replace the compact placeholders, with nominal 345 ×168 mm unit references, fitted gas branches, two open exhausts and four supports/insulation clearances. Detailed proportions and installed poses remain estimates; flight wiring/ICDs unresolved. All 740 tests pass; useful real/model comparison and installed desktop/phone samples saved. Ships with the 32-part numbering/visible-surface anchor pass below; part 5 stays grouped. Published 92ef530; Pages 40734789253 succeeded. The merged project passes 746 tests; concurrent project files preserved unchanged. Live 32-entry list, 33 selector options, grouped navigation and part24 verified without overflow/errors. Public verification.

October 27: Complete inspection numbering and anchor check. All 32 selectable assemblies have matching menu/canvas/list IDs; original 1–24 preserved. Pins use actual visible mesh surfaces and follow motion/cover changes. Part 5 stays grouped. 733 tests and desktop/phone checks passed before the blower replacement above; both passes are included in the current publication.

October 27: MEB harness connection and isolation. The curved line by part 5 now seats in an estimated external connector shell and strain relief; the same single branch remains visible with part 5/MEB or Bay harness. Actual flight connector/circuit unknown. All other routes and enclosures preserved. 725 tests and desktop/phone inspection checks pass; archive and samples saved. Published 00a7e35; Pages 40625667490 succeeded. Concurrent shared-project work preserved unchanged; 729 combined tests pass. Live MEB isolation, 24 parts, Open interior and no overflow/errors verified; public sample saved.

October 27: DrACO electrical evidence and blower scale audit. Historical audit: real supplier photos and BNCP-74-000 specifications revealed that the then-current blower pair needs a coordinated housing/placement/plumbing rebuild. Added a nested, closed evidence panel with inspection links; no speculative sensors or wiring added. 721 tests and desktop/phone checks passed. Its compact-pair measurements/preview now load the archive; the October 28 replacement above supersedes its next step. Published 5e25b65; Pages 40602293824 succeeded. Live build, closed audit, Open interior, 24 numbered parts and no overflow/errors verified; public sample saved.

Open the simulator

A browser-based, NASA Dragonfly-inspired flight and telemetry demo. Includes a 3D rotorcraft, continuous Titan terrain, manual flight controls, and fixed and 360-degree free cameras. This is an educational simulation, not an official NASA flight product or a validated flight dynamics model.

New to the code? Start with ARCHITECTURE.md (file map, shared state, house rules) and the latest entries in CHANGELOG.md.

October 27: Fixed feed lead and numbered open-interior defaults. One estimated stationary feed-drive service lead per drill, with a real cover opening/grommet and supported local retention. Flight circuits/remote routes remain unknown. Component selector numbers match the 24 model labels. Fresh/reset inspection defaults Open interior; fresh Mission uses Internal, preserving explicit saved modes. 712 tests and desktop/phone checks pass; sources, archive and samples saved. Published 6ee178c; Pages 40596670874 succeeded. Concurrent shared work was preserved unchanged and all 717 combined tests pass. Live Internal/Open default, 1–24 numbering, reset behavior and fixed-lead disclosure verified; public samples saved.

October 27: Upper drill drive fit. Corrected the exposed screw-to-drive drawing gap and upper mount crossing; added supported motor seating, panel openings and hollow coupling cover/fittings. Hidden joint/gears remain unresolved; exposed shaft envelopes are explicitly illustrative. Carriage, leads, bellows, hoses, travel and bit depth preserved. 682 tests and desktop/phone checks pass. Next: fixed feed-actuator lead evidence. Published badbec2; Pages 40591034163 succeeded. Concurrent shared-project work was preserved unchanged, and all 703 combined tests pass. Live build/disclosure, 24 parts/two targets and no overflow/errors verified; public sample/JSON saved.

October 27: Drill roller carriage and spherical nut. Six crossed roller exteriors meet a V-shaped guide; an open-bore spherical nut/pocket follows real ETU/EM1 hardware. Roller arrangement uses a labeled engineering drawing; fitted dimensions and finish remain estimates. 675 tests, desktop/phone/layer checks pass. Drive covers, sleeves, cables, hoses, travel and bit depth preserved. Next: upper drive coupling/support reconciliation; hidden internals and wiring remain open. Published ee24ad0; Pages 40505971224 succeeded, including all 675 tests. Live build, carriage disclosure, 24 parts/two targets and no overflow/errors verified. Public sample/JSON saved; preview stopped and viewport restored.

October27: Feed covers and paired dust sleeves. Real EM1 photo-informed cover/gearbox, metal guide finish and moving bellows; dimensions remain estimates. Covers can expose the screw and provisional actuator.130 mm travel, bit depth, hoses and moving leads retained.669 tests and desktop/phone checks pass. Next: roller carriage/self-aligning joint; exact flight configuration remains open. Published dc0ee8a; Pages40497034993 succeeded. Live build, stage/limits text, 24 parts/two targets and no overflow/errors verified. Public sample/JSON saved; temporary preview stopped. Previous release entries below are historical.

October27 wrap-up: Moving drill lead examples. Both drills now have constant-length lead loops between moving motor caps and fixed local supports. Their flight routing/material remains unknown.662 tests, desktop/ phone Internal/Thermal checks pass; mechanics and sample plumbing preserved. New real EM1 and related TRIDENT source photos saved. Published b6fb179; Pages40481602061 succeeded. Live build, new disclosures,24 parts/two drill targets, no overflow/errors verified; public sample/JSON saved. Local preview stopped. Next pass is the2024 feed-stage/gearbox/bellows reconciliation; checklist has exact handoff.

October27: Both-drill plumbing trace. Both intake routes exist through the shared carousel/return. Corrected a1 mm gap at the blower selector; only the final return control advances4 mm into a3 mm EST seat. Collapsed inventory explains the route; valve interiors remain unmodeled.655 tests pass. Published b27cb6e; Pages40475925194 succeeded. Public build, full-network focus, route/limits/seat text,24 parts/eight rows and no overflow/errors verified. Public sample/JSON saved; temporary previews stopped. Previous release details are historical.

October27: DrACO side-by-side equipment audit. Both drill mechanisms retain matching 45-piece inventories. Separate port/starboard inspection targets have pinned numbers; sample hoses/tubes hide as plumbing rather than electrical harnesses. New evidence/coverage notes stay collapsed. All hardware geometry is unchanged; complete wiring and exact flight blower locations remain open. 652 tests and desktop/phone checks pass, with 24 numbered parts and 11 IEM targets retained. Published 4f375be; Pages40463472979 succeeded. Live build, eight inventory rows, both drill focuses, 24 parts and 11 IEM targets verified without browser errors. Public sample/JSON saved; local preview stopped. Previous release below is historical.

October26 cleanup: IEM tag and terminal wraps. Lead8 label no longer clips; terminal blocks become hollow wraps. All26 cable curves, endpoints/radii and other fittings remain exact.645 tests; desktop/phone retain24 parts and11 focuses. Literal terminal capacity mismatch remains disclosed. Published 7538184; Pages40278240471 succeeded. Live build/disclosures,24 parts and11 focuses verified, zero browser errors. Public sample/JSON saved; preview stopped.

Previous published build 020bfae: all eight IEM columns. Final five leads plus approved4/13/21 refits; other18 installations and26 endpoints/radii retained.640 tests pass, all325 cable-pair screens pass; desktop/phone retain24 parts. Measured return extension10.4 mm fits the approved2 cm allowance. Terminal convergence and an inherited lead8 own-tag contact remain open. Pages40276054426 succeeded; live build/status,11 targets,24 parts and phone375/375px verified, zero browser errors. Public sample/JSON and recovery tag saved; preview stopped. Git cleanup remains paused.

Previous published build 00d32be: IEM columns five/six and two-finger inspection pan. Seven estimated routes/tags/supports and fourteen ties; all prior routes and endpoints retained.21 reviewed,5 open.635 tests pass; desktop/phone checks retain24 parts. Git cleanup remains paused at the unverified backup; no deletion or game-data changes. Pages40241403099 succeeded; live build/new focuses/disclosures and phone375/375px verified with24 parts and zero browser errors. Public samples/JSON and recovery tags saved.

Previous published build a68e08a: right-drag inspection pan. With Rotate selected: left drag rotates, right drag pans, wheel zooms. Shift+arrows also pan; focus/reset recenter.628 tests pass. Hardware and flight are unchanged. OneDrive warning handoff and command-local no-cleanup Git flags are recorded there. Pages40222640031 succeeded; live pan/zoom/24-part checks pass without browser errors.

Previous published build 051c9d9: inspection controls. Rotate-selected mouse-wheel zoom; layer selector above the viewport; Interior c adds Thermal's grey grid. All controls/details retained, with corrected phone section order. 623 tests pass. Flight physics and all hardware are unchanged; IEM15/16/17 remain next. Pages40217478630 succeeded; live wheel/grid/layout/24-part checks pass without errors.

Previous published build e372a1b: IEM fourth-column leads 12-14. Three estimated routes with attached labels, supports and ties; eleven completed routes and all endpoints stay fixed. Fourteen reviewed, twelve remain; 619 tests pass. Shared reference imagery is reused, with one current model sample instead of another comparison. Pages40213913093 succeeded; live target/count/source/24-part checks and desktop/phone render checks pass, zero browser errors. Recovery tag and single public sample saved.

Previous published build 2491558: IEM third-column lower pair, IDs 10/11. Two source-labeled estimated routes with attached labels, supports and lacing. All nine completed routes and 26 endpoints stay fixed. Eleven reviewed, 15 remain; 614 tests pass. Real/CAD/model comparison and full-IEM desktop/phone samples saved; terminal overlaps stay open. Pages run 40211931773 succeeded; six live focuses, five source links and all 24 parts verified with zero browser errors. Public full-IEM sample and acceptance record saved.

Previous published build 09e58eb: IEM third-column upper pair, IDs 8/9. Two estimated service routes, attached labels, supports and lacing; seven completed routes and all endpoints stay fixed. Nine local routes reviewed, 17 remain; 609 tests pass. Full-IEM samples retain all columns; subassembly isolation never deletes model geometry. Pages deployment 40171467980 succeeded; the live build, five inspection focuses, source links and all 24 numbered parts are verified with zero browser errors.

Previous published build 6933908: IEM second-column upper/lower routes. Both routes, labels and their own fittings are refined around the unchanged middle lead. All 26 endpoints and unrelated geometry stay fixed; 604 tests pass. Seven local routes reviewed, 19 still open. Terminal bundle overlaps and actual flight circuits remain unresolved.

Previous published build 95d5f89: IEM middle-lead route. One overlapping lead now has a separated estimated service bend, attached label, support and two ties. All 26 endpoints and other geometry stay fixed; 599 tests pass. Five local routes reviewed, 21 remain; the inherited terminal bundle overlaps are explicitly open.

Previous published build 31afae8: five-family installation pass. IEM/PSUs/radio/battery gain 10 estimated supports and 6 lacing examples; navigation gains retention details without invented wiring. Collapsed status panel links all five families. All prior geometry and 62 curves unchanged; 595 tests pass. IEM middle support and complete flight routing remain open. Real/CAD/model samples, register and recovery archive saved.

Latest published installation: IEM lower connector pair, build38f297e. Four leads now have reviewed installation estimates;22 remain open. Lower bends, attached labels, lacing/supports and a third focus are added without changing the upper pair, enclosure, floor wiring or any of26 endpoints.583 tests pass; source/CGI/model samples saved.

Latest published installation: IEM upper service pair, build19b379f. Two existing leads gain estimated slack, lacing/supports and separate focus access; all26 endpoints, enclosure, main harness and other24 paths remain unchanged. Real/CAD/before/after samples and recovery archive saved.580 tests pass; full IEM/attic wiring remains open.

October24 continuation: sampling electronics and unresolved interfaces. New collapsed Mission section separates lander electronics, DrACO avionics, DraMS main/ laser electronics and warm drives, with primary-source links and four existing-model focus shortcuts. No new wire/model or assumed bus. Actual flight interfaces remain open.

October24: carousel warm-drive installation examples. Three existing drive locations gain photo-inspired casing/cap/mount detail, individual focuses and separate layer controls. Estimated local leads/supports stop before any unmodeled bus. Cold carousel and neighbors stay fixed. Real/CAD/before/after samples saved;572 tests pass. Full attic wiring and actual drive/connector identification remain open.

October23 attic continuation: local-line register and DraMS wire installation. Electrical examples, gas/sample tubing examples and unclassified routes now hide separately for DraMS/carousel. One short wire gains estimated slack/lacing/support with fixed endpoints; other hardware/tubing/carousel geometry unchanged. Full attic trunk and drive/feedthrough interfaces remain open, not inferred from integration-test cabling.562 tests pass. The interior is being redrawn part by part from sources, tracked in the Interior realism checklist; every photo, paper page and patent behind each part is listed in research/PART-SOURCES.md. The part numbers in Internal and Thermal now stay on their parts as the view turns (October 15); a crowded number moves to a nearby spot with a short leader line back to its part.

October23 Codex: attic/local insulation and Open interior adds removable attic/HGA surfaces, estimated lower-camera window surrounds and an access preset without enclosing walls/foam. Camera windows now meet the actual beveled body; camera bodies/aim remain fixed. Information/source tabs updated. Nested Three.js core now has a release cache key. Full foam construction and flight harness topology remain open. Next: endpoint/harness coverage audit, then one supported connection/support family at a time, including the missing attic trunk.

October23 continuation: bay endpoint audit and lower-camera installation. Measured20 branch/network-link rows; closed three lower-camera gaps and added the port Down Camera bypass around PSU1 through existing reliefs, without moving equipment. Routing/support estimated; unnumbered Bay harness has four focuses and wire/mount masks. 24 numbers retained; full attic and other networks remain open. Earlier public59a03b9 eventually initialized normally; genuine Open interior evidence is saved in new research.

Earlier October22 Codex: removable body-insulation surfaces adds top/deck, port, starboard, belly and end-region inspection masks. The base outer shape, 24 equipment parts and structural panels remain unchanged; old cross-bay seam blocks are archived. Real NASA/APL foam test photos and NASA NESC development slides are saved. Regions are surface-only estimates, not flight tile cuts or actual foam volumes. Next: attic regions, then local camera/HGA insulation; legacy thermal assumptions unchanged.

Earlier October22 Codex: top deck / attic edge adds separate faces, a gimbal/dome clearance opening, short attic-edge lip and estimated stow-lock foundations. Three focus targets and independent masks; HGA/TWTA/shared mounting plane and original24 parts unchanged. Core/layup and local construction remain estimates. Next: body/deck insulation-region separation, then attic foam and local camera/HGA regions.

Earlier October22 Codex: port wall and end joints extend the construction view with separate port camera reliefs and estimated nose/canted-aft joint pairs. Select Port structural wall or Wall end joints (estimated) for their targets and independent masks. Original24 numbered parts and starboard geometry unchanged. Next: deck/attic boundary, then separate foam regions and local camera insulation.

Earlier October22 Codex: first structural side wall adds separate aluminum faces and edge fittings, with NASA's published 0.254 mm face-sheet thickness. Select Starboard structural wall for its three targets and independent masks; the original 24 parts remain unchanged. Outline, panel separation and fittings are estimates. Opposite wall, end joints, deck and separate insulation regions remain queued.

Earlier October22 Codex: MMRTG generator exterior now follows NASA's real generator photo: pale thin fins, narrower housing, root attachments and stepped metal covers within the published64 x66 cm envelope. Detail/finish remain estimated. Select MMRTG generator body in the existing inspector for housing/fins/ends and separate masks;24 numbered parts stay unchanged. Sensor21 is refitted at actual scale. Next: first structural side wall, then the extensive connections/layer/UI roadmap.

Earlier October 21 Codex: fin-root RTD example completes the last numbered initial pass: 24/24 (9 Codex / 15 Claude). Real Omega catalog photos/drawing guide a temperature-appropriate family approximation with separate layers and macro inspection. Actual flight package/installation remain unconfirmed. The MMRTG body, structural wall layers and extensive connections/UI roadmap are still open, in that order.

Earlier October 21 Codex: battery-temperature sensor example replaces part20's annotation with a small TE thermistor-family approximation, estimated bonding and paired leads. Real manufacturer photographs and drawing archived; actual Dragonfly package unconfirmed. Bead close-up and independent layers preserve physical scale. 23/24 initial passes;21 remains an annotation. The checklist now details missing connections, construction layers, view presets and the final whole-project UI/source audit. Next:21, MMRTG body, side wall.

Earlier October21 Codex: RDE connector installation adds estimated slack loops and lacing to the four existing leads, with a wire-only Harnesses mask independent from connector Interfaces. Claude's casings and main trunk remain unchanged. At that checkpoint the count was22/24; other units' loops/installation remain open.

October 21 (Claude): main wiring harness, first pass: silver bundles on the bay floor, a centerline trunk with branches to the units, after NASA/APL's September 2026 wiring photo (routes estimated). October 20: photo-detail passes on the Wonderwall and DraMS, and the arm shoulder fittings.

October 20 (Claude): part 21 MMRTG fin-root annotation replaces the green sphere that passed through the generator's top fin with a labeled target on the housing beside the fin's root, and shows the MMRTG when part 21 is isolated. Like part 20, it is not a reconstructed flight sensor; part 21 remains open.

October20: part20 battery-temperature annotation replaces the intersecting green sphere with a labeled measurement target and keeps battery context during inspection. It is not a reconstructed flight sensor; hardware evidence and part20 remain open. Battery geometry, part21 and simulation behavior are unchanged.

October19: part09 evidence and inspection update adds individual focus for the two provisional rotor controllers and removable covers/casings/ mounts/interfaces, without changing their estimated geometry or any flight physics. Clearer hardware imagery is still needed;09 stays open and the initial-pass count remains20/24.

Swarmfall: Drone Invasion (game)

October 20 — FIX-07: phone-sized screens collapse Objectives and Battle log after the first capture. Tap either heading to reopen it; that choice stays respected during the mission. Captured saves resume compact on phones. Warning toasts remain active. Automated checks and acceptance limits.

Play Swarmfall (its own page, dist/invasion/; the simulator above is unchanged). A strategy game built from the simulator's best parts: drop drone swarms on planet after planet in a campaign galaxy and take each world by force or by quiet control.

• Campaign: seven planets invaded in order, from Titan (an empty practice planet on the simulator's real Ahmakiq Undae terrain) through rising society levels (tribal, agrarian, industrial, networked, spacefaring) and threat levels. Each win opens the next planet and pays matter, new drone types or bigger drop pods.
• Drops: a squad arrives as a wave of parachute pods carrying 1–3 drones each; a large Dragonfly travels alone. Slot capacity, matter cost and recharge apply once to the whole wave. The drop panel shows pod count and courts before launch. Pick a zone, inspect the spaced court outlines, then choose Launch here. Unsafe/occupied courts refuse the drop without charging matter. Drones arriving later retain their squad's orders. Held zones pay matter over time and become landing areas. Existing paid single-pod saves keep their original payload.
• Damage or control: Strikers and Breachers destroy defenses (salvage pays matter, but the public's alarm rises); Links capture under fire, calm the public and turn standing defenses to your side. High alarm brings militia raids, stronger defenses and artillery.
• Hazards: methane rain, gust fronts, dust storms, acid rain, blizzards, ash clouds, eruptions, lightning and solar flares, each announced before it starts. Zones you hold give shelter.
• The Dragonfly is an unlockable drone: recover NASA's lander from its landing site on Titan and it joins every later drop (the simulator's research model, interior removed).
• Objectives and achievements: two optional goals per planet (three stars) and twelve campaign achievements, saved in the browser. Required captures offer Leave for victory or Continue capturing. Continuing keeps a visible extraction action and remaining-zone count. Earned victory remains available if zones are retaken, but total defeat can still lose the mission. Stars and time goals finalize on extraction. Recalled survivors refund half their cost; unreleased pod cargo is fully refunded. Each attempt's result settles once. Watchtower/jammer goals explicitly allow destroy or capture, with a live hostile count. A defense retaken by the enemy counts against that goal again until neutralized.
• Mission saves: active battles save on this device every five seconds and when the page is hidden or left. Choose Resume mission on the title screen; the restored battle starts paused. Starting another mission asks before replacing that save. Browser storage must be available; campaign progress and the active battle are saved together. Resume retains active/inbound squad selection, camera, draft payload and drop-panel state. It shows the HUD at 1×, cancels landing placement and requires a fresh explicit launch. Existing issued orders and paid cargo remain in the battle; resuming adds no purchase. Earlier mission/progress saves migrate automatically. Results save immediately with the completed checkpoint cleared; a storage failure shows a retry action and retains the result in memory. If storage stays unavailable, keep the page open until it can save. resume validation now checks unit/order/zone references, complete pod waves, combat/hazard fields and bounded data before restoring. Old v1/v2 checkpoints remain supported. on the title screen now offers session/original JSON downloads, preserved browser backups and retrying storage access. Unsupported/damaged saves show a recovery notice; recovering readable campaign data or starting fresh requires confirmation and a verified exact-original browser backup before replacement. Failed backup or writes retain the original. Browser backups share this device's storage; download a file for an external copy. Real browser/download/touch acceptance remains open.

All game numbers, the invented worlds and the drone and defender models are gameplay choices, not physics. Rules are in plain modules with tests (tests/invasion-*.test.mjs); see ARCHITECTURE.md. Later ideas from the owner: a free mode with a galaxy of planets to pick from, and a local versus mode where players race to take the same planet with drone-vs-drone weapons. The Swarmfall prioritized update checklist was expanded October 19 with ordered milestones, current implementation status and detailed acceptance tasks. Multiple parachute pods and victory/continue remain first. It also tracks playable cities, destruction, pilot control, drone-specific tactics, better graphics and a separate persistent free-roam planet mode alongside the campaign; these are planned features. The RTS/space design study links reference games and proposes connected territories, scouting/relays, recoverable drones and modular loadouts. Existing task IDs and the full previous checklist are preserved. Future visual priority ART-09 reviews and upgrades small assets that lag behind the best existing work, including unit details, props, icons, textures and effects, while preserving originals and keeping performance and gameplay readability in view. Scout's first hull/sensor polish pass is implemented; before/after and checks record the preserved original and pending real-browser/phone art review. Inspect it in the lab: Original Scout / Polish pass. Both offer orbit, zoom and turntable controls; the version links reset the camera.

Asset & Environment Lab

Open the public Asset & Environment Lab. Choose Environments for the five city tests and their four destruction stages, or Assets to inspect 18 individual models: four Swarmfall drones, five urban buildings/props and nine fantasy pieces. Each asset has creator/license details, orbit/zoom views and a turntable. The lab is also linked from the game's title screen. It is separate from campaign progress. October 19 — Codex: damaged city stages now include animated smoke, flame tongues, embers and warm local firelight. Major destruction burns more intensely; ruins have smaller residual fires. Use Smoke, Fire & embers and Pause effects independently, or Light quality for fewer particles and no added point lights. Reduced-motion preference starts the effects paused. Compatible individual architecture assets also offer destruction previews; the intact originals and drone inspector remain available. Try Skyway with smoke/fire or the alien arcology. These are visual aftermath studies; spreading fire, combat damage and sound remain future work. Implementation and verification. October 17 publication supersedes the earlier local-preview-only status recorded below; all existing local links, historical foundation versions and original asset files are retained.

The separate Asset & Environment Lab compares city directions without changing campaign planets or mission saves. Run python research/fantasy-city-2026-10-16/preview-server.py from the repository root, then open the local gallery. It serves the gallery on port 4201 with fresh resources; the existing game preview can stay open. The first three tests are Alien arcology, Metropolis and Tidal citadel. Two complete Blender-scene tests now supplement those original compositions: Cyberpunk City / Skyway megacity and drone City / Phototextured city. Skyway is the strongest intricate/alien architecture comparison; Phototextured City provides the more realistic urban comparison. Both retain the source city layout and use a normalized 2 km span for inspection, not verified real-world dimensions. Cyberpunk City's new land follows individual architectural contact points, with small footing pads, coastal slopes, access roads and animated ocean water. Foundation frames the ground; the credits panel links to the retained flat-ground and first two foundation passes. The owner prefers the second low-coast pass, retained as the baseline; fitting all elevated bases remains open. The later high terrain envelope was rejected because it swallowed buildings and is archived. Foundation tooling shortlist records Blender, Houdini and Unreal options for future fitting. Skyway megacity and Phototextured city remain the gallery titles, with original asset names in credits. These are local previews, not a published update. The Blender fitting test preserves the earlier underside-fitted structural-fill experiment on that second coast. Blender authoring project and tool status record the reusable Geometry Nodes setup, portable Blender and verified SideFX Launcher download. Houdini 22.0.466 Production is now installed with Apprentice licenses activated (October 17). The owner rejected the building-extension approach and requested a Houdini terrain fit. The is the current Skyway preview. It uses a connected landscape, projected building bases and constrained terrain steps, with the second coast's outer shore and animated water. and add inspection angles. record the native node graph and independent gap/burial checks. Source buildings are unchanged; the earlier platform, second coast, flat-ground and Blender passes remain available. Terrain art direction still needs owner review, especially the high rear districts in the source model. Each offers vista, district and street cameras, orbit/pilot movement, flyover, dusk and a lighter render setting. These are visual studies; collision, campaign navigation, streaming and destruction remain checklist work. identify the imported CC0 models and original compositions.

The reusable source library is C:\Users\Heemi\Documents\Game Asset Library, outside this repository and available for other projects. Six completed backups total 1,843,813,552 bytes: cyberpunk City, drone City, Downtown City MegaKit Standard, Kenney Fantasy Town Kit, Spaceship HY-21 and Poly Haven Hidden Alley. Its README and JSON catalog retain source links, creators, licenses, SHA-256 hashes, archive inventories and pending downloads. Backup status and handoff. All existing work, original archives and Downloads copies are retained; no paid packs were bought. The first gallery's exact source and screenshot are preserved in the v1 archive.

The visual target remains large, intricate, less blocky alien/fantasy cities. Minecraft is a scale/infrastructure reference; Skyrim, Spider-Man, Crackdown, Kakarot and Synty Elven Realm remain in the city research and complex-scene shortlist. Spacecraft, orbital stations and planetary/alien architecture are priorities in the download queue. Future individual-drone control (MODE-20, PILOT-01–09) includes switchable first-person onboard and third-person chase views, flight/aim/abilities, and a return to RTS command while other squads keep their orders. An optional dedicated pilot/skirmish mode can come first; campaign integration remains the eventual goal. The owner's Ghost Recon / RTS reference guides the idea; the exact title is unspecified. The later Tank Wars-style builder (MODE-21) also remains planned. Victory extraction/continued capture and multiple pods carrying 1–3 drones each are now implemented, with browser/device acceptance still open.

October 17 — progressive destruction artwork: all five city tests now offer Intact, Damaged, Major destruction and Ruins. Broken facades, jagged tower sections, exposed floors, bent rods and settled rubble are actual derivative 3D meshes; the original architecture, terrain, ocean and citadel foundations remain available. Start with the alien destruction test and switch maps/stages in the sidebar. These are visual prototypes; battle damage, collision, save persistence and collapse animation are still open. Game integration review and all five test links record the measured asset sizes and concrete route into campaign districts. Multiple-pod delivery remains the next requested gameplay pass; parachutes were not changed here.

Run Locally

Serve the dist folder over HTTP, then open the local URL (the game is at /invasion/):

python -m http.server 4176 --directory dist

The app uses JavaScript modules and WebGL 2. Dependencies are included locally; there is no package installation or build step.

Checks

node --test tests/*.test.mjs

Publishing

Pushing to main runs the checks and publishes dist to GitHub Pages. Local files reference each other with ?v=dev. The publish workflow replaces every stamp with the commit hash and writes the build into the header ("Build 1a2b3c4 · date"), so phones always fetch fresh files and you can see which version is live. Leave the ?v=dev stamps as they are; a test checks them.

Record every change in CHANGELOG.md, signed with who made it. The older Sites copy is managed separately and is being retired; sites-redirect/ holds the page that forwards its visitors here.

Assets

The October 10 NASA/APL source review records eight official references, including NASA's printable Dragonfly model, and proposed realism changes. The October 11 inspection update implements the reference viewer, selected geometry corrections and component diagnostics. APL model reference assets and credits preserve the originals; compact viewing copies load on demand in the Mission view.

Three.js 0.180.0 is included under its MIT license. Inspection toolbar icons are from Lucide, with their license included. The Titan terrain reference image was supplied for this project. NASA mission references are linked within the app.

Landing Area

The landscape follows the published description of Dragonfly's landing area, Ahmakiq Undae (named by the IAU in September 2026): dunes and interdunes south of Selk crater, reaching to the edge of a range of hills. The base sits in a flat interdune corridor about 2 km wide between two long dunes 75-120 m tall that run roughly west to east, with hills beyond the northern dune. Dune sizes follow published Cassini results; the exact layout is illustrative. Dune sand is dark and pebble-free; the interdune keeps the Huygens-image ground texture. The previous maps are archived in archive/.

Flight Controls

• Pilot fits the flight box to the available browser height while keeping telemetry, the flight graph, science, mission controls and references on the scrollable page below. The small fullscreen button expands only the flight box; press it again or Escape to exit. Native browser zoom remains available. In Free camera, drag to orbit and scroll, pinch or use the + / − buttons to move the camera in (to about 2 m, close enough to see the antenna's slots) or out; Fixed always uses the normal chase distance. The reset icon beside Diagnostics requires a second tap to confirm; the footer also retains Start over from the landing.
• Takeoff climbs to 40 m and hovers. Cruise holds altitude (at least 20 m) and flies forward at 10 m/s. Land descends at up to 1.3 m/s, slows near the ground and idles the rotors after touchdown. Auto resumes the demo profile from the phase that matches the aircraft (takeoff if landed, hover if airborne).
• The throttle is a climb-rate command: 50% holds altitude, higher climbs, lower descends. By default it is sticky and stays where you leave it. The "Throttle: sticky / centering" toggle under the stick makes it spring back to 50% on release. Yaw, pitch and roll always spring back to center.
• Sticks respond to how far you drag from where your finger lands, so touching a pad never jerks the controls. Moving the throttle cancels a button's altitude hold; steering does not.
• Attitude, climb rate and forward speed are rate-limited (climb 3 m/s, sink 2.5 m/s, about 1 m/s² horizontal), so mode changes never teleport the aircraft. The chase camera eases between Fixed and Free instead of snapping.
• Keyboard: W/S throttle, A/D yaw, arrow keys pitch/roll, Space levels off and holds altitude.
• Reverse (sim) beside Throttle: sticky arms a tiny fictional reverse-thrust brake, off by default. In manual flight, lower the throttle all the way to 0% above 1 m: the pulse adds up to 0.2 m/s² of downward acceleration, continuing through the climb apex until reaching 0.6 m/s descent. It uses a small modeled rotor pulse, reversed rotor animation and 50 W extra electrical load [EST]. Normal descent speed and the gentle touchdown flare stay unchanged. It does not operate in Auto, flight plans, altitude hold, safety descent, arrival or over liquid. Both views show off / armed / down thrust, and landed saves retain the enabled setting. This is a simulator convenience, not a capability of the real Dragonfly's fixed-pitch rotors.

Vehicle Inspection

Mission's Internal and Thermal layers have an inspection toolbar for the research model. Choose a component to frame it, optionally isolate its assembly, zoom, or enable drag rotation. With the canvas focused, arrow keys orbit, +/- zoom and Home resets the view. These inputs do not steer the vehicle. Native Ctrl browser zoom and page scrolling remain available; touch rotation is opt-in. The original mock-up keeps its existing view.

Follow sampling tracks the shared drill, transfer, carousel and DraMS stages. Manual component selection ends follow mode. The pause icon pauses the whole simulation, so the science clock, power and sample results cannot drift apart; the inspection camera remains usable. Reset restores the whole model. Geometry, stage timing and visual flow markers are illustrative, not flight CAD or onboard imagery.

The expandable Inspection layers section hides the body/attic shells, separately modeled foam, compartment panels and flow markers. Component casings, covers, mounts, interfaces and harnesses can be hidden independently where modeled; the component menu marks hidden settings. Restore layers keeps the current camera and selection; the camera reset restores everything. These are viewing masks on existing geometry, not a verified structural layer stack.

The two EFIELD probes follow real 2025 engineering-model photos, with published 30 cm carbon-fiber stalks and 5 cm gold electrode balls. Preamp size and exact installation remain estimated. The METHAN cell is an owner-approved, labeled design approximation with removable filtered covers, supports and fiber entry; it is not photo-matched hardware. Warm DPU sources/detectors and hidden optics are not invented. The subassembly menu focuses Front EFIELD, Rear EFIELD or METHAN independently. Initial asset pass #22 is complete; the current count is 20/24, including the later approved navigation approximation. Geometry: Codex.

The nose bulkhead study replaces the old block with a provisional thin face, perimeter returns and mounts. Real integration photographs, separately labeled CAD and process renders are saved together. Complete nose-face geometry and dimensions remain unresolved, so #03 stays unchecked; the current total is 20/24. Hiding the face shows unmodeled space, not a verified hollow flight panel. Geometry: Codex.

The part-05 approved navigation approximation replaces the plain Ocellus boxes with estimated machined MEB structure and FEB base/returns, receiver barrel and angled collimator housing, retaining NASA's nominal 2023 envelopes. Real related-hardware photos inform construction; identified CAD supports scale/layout, not actual finishes. Both Honeywell space MIMU family models and every installed pose are unchanged. Four unit focuses plus shelf, independent casings/covers/mounts/IMU interfaces and original05 anchor retained. All units clear rebuilt neighbors without moving them. Hidden boards, detailed optics and harness remain unmodeled. #05 initial pass checked as an owner-approved approximation, not flight verification: 20/24 total (5 Codex,15 Claude). Exact flight variant/configuration/interfaces remain follow-ups. Geometry: Codex.

The later Ocellus evidence follow-up saves original NASA SQRLi review slides/hardware photos, identified Ocellus packaging CAD and a credited real digitizer-test photo. They improve the reuse/layout evidence, not flight-box photo verification. Research only: the published IMUs/lidar studies and app behavior remain unchanged in that historical checkpoint. The later approved outer-structure pass above supersedes its unchecked/count19 status; the evidence limitations still apply. The deeper cross-project search traces predecessor lidar electronics, SpaceCube Mini, real fiber-interface heritage and supplier leads back to Ocellus. Credited original pages/images and an evidence comparison are saved. Confirmed family reuse is separated from unverified scanner/processor variants; no unrelated project's hardware or hidden boards were installed in the model.

The aft bulkhead follows NASA/APL's real structure photos: gold-colored face, ribbed circular generator interface and pocketed rim. The generator rises aft at an estimated 25 degrees; the fan collar and aft body profile follow the same interface. Exact dimensions, coating and internal air passage remain unverified. The source fixture is explicitly inverted in NASA/APL's caption. Geometry: Claude; closeout: Codex.

Model assumptions now separates sourced design references, approximate geometry, invented training geography, simplified engineering and deliberate simulation additions. Implementation and limitations: interactive inspection notes.

Arrival at Titan

On a browser's first visit the app opens with a one-minute, time-compressed entry, descent and landing sequence following the published timeline: entry at 1,270 km, drogue and main parachutes, heat-shield separation, the lander lowered below the backshell, rotor spin-up, release at 1,000 m and powered flight to touchdown. Skip ends it; "Replay the arrival at Titan" at the bottom of the page plays it again. The hardware sizes (4.5 m 60-degree aeroshell, 8.25 m disk-gap-band drogue, 16.7 m ringslot main) are published; the textures, plasma glow, sparks, smoke puffs, haze wisps, touchdown dust and camera moves are artistic. Under the main parachute the camera looks down past the canopy while the ground emerges from the haze, and it closes in on the lander within about three seconds of release.

Sound

Sound is off until you turn it on with the ♪ button next to Mission and Pilot (browsers only start audio after a click or tap; the choice is remembered). Everything is synthesized, with no audio files: the rotors' low chop at the blade-pass frequency (rpm / 60 x 3 blades, 35 Hz at 700 rpm) plus a faint motor whine, wind that rises with wind and flight speed, and in the arrival the entry roar, parachute flutter and a thump at each separation. Pitches follow the simulation; levels and textures are artistic. Sound pauses when the page is hidden.

Sound on Titan. The Huygens probe measured the speed of sound at the surface at about 194 m/s (Earth: 343 m/s), so sound takes about 1.8 times longer to arrive: in the arrival the separation thumps are heard up to about a second after you see them, depending on how far the camera is. Titan's air is 4.4 times denser than Earth's, so the same vibrating part radiates about 2.5 times the sound pressure (about +8 dB), and wind of a given speed is noisier; the wind sound is weighted for that. Pitches set by rotation (rotor chop, motor whine) are the same as on Earth; pitches set by air resonance (voices, pipes) would drop to about 0.57 times, but none are simulated. A sand hiss joins in when the rotor wash raises dust.

Rotor-Wash Dust

Near the ground the rotor wash lifts sand. In hover the wash is about 4.4 m/s, calculated from the published mass, Titan's gravity and air density and the rotor size, and sand starts moving at about 1 m/s. Dust appears when the rotors are working within about 6 m of the ground: more with more throttle or a hard climb, swept behind in fast forward flight. It rolls outward from under the rotors, rises and fades, and stays where it was raised as the vehicle moves on. Dune sand gives the most, the rain-dampened ground and the icy outcrop much less, the puddle none. The arrival keeps its own touchdown dust. Reach, rates and the look are estimates.

Vehicle Model

The default 3D model follows the NASA/APL 2023 thermal drawings, with details refined from APL's 2024 CAD visualization and NASA's 2026 printable-model assembly guide: a long insulated fuselage, rounded forward "attic", MMRTG between two tail fins, four arms with coaxial three-blade rotors on machined motor stacks (drawn from APL/NASA's "Revolutionary Rotors" wind-tunnel video), flat landing runners, and a high-gain antenna disc that rises for downlink on its motorized arm and azimuth/elevation gimbal. The Model button in the vehicle view switches to the original demo model for comparison. Proportions are measured from the drawings; the overall size matches the published 3.85 x 3.85 x 1.75 m. Rotor blades are tapered and twisted with a thin cambered section and rounded tips, with hub caps, and spinning rotors show a faint blur disc that strengthens with rotor speed. Blade chord, twist and thickness are estimates from the drawings.

Under Mission > Vehicle Layers > NASA / APL references, the payload cutaway and rotating CAD movie retain visible credits, source links and design dates. Compact local copies load only when opened, and the movie pauses when closed or when switching to Pilot. NASA's STL/3MF files and assembly/print guides are linked with Gary Allen Backus / Backus Creative Imaging credited. The simulator still uses lightweight procedural geometry, not imported flight CAD or the large print meshes.

The updated model has body-side DrACO drill/feed assemblies, estimated pneumatic tube routes, flanged rotor-arm roots, separate coaxial motor housings and broad skids with shaped supports. Internal and Thermal distinguish the IEM, two PSUs and Frontier radio. Selecting a component shows relevant live simulation readings and primary-source links. These share the existing thermal/system model; there are no invented independent processor or PSU sensor readings.

First Expedition

Begin expedition starts a six-step science exercise across three distinct sites: collect a dry-sand baseline at base, scout the rain-darkened interdune and return, fly back to sample it, sample the ice-rich outcrop, return to base, and downlink the records. The objective explains the next step and its scientific purpose. Prepare flight creates a suggested 40 m route in the existing planner: review its GO / NO-GO checks, then explicitly uplink it. Use 5x / 20x during flight and up to 100x during downlink. Recharge or wait for daylight in Diagnostics when necessary; the expedition does not bypass the energy, thermal, scouting or radio restrictions.

Science notebook and debrief is expandable in Mission and Pilot and keeps one new sample from each expedition site, including its location, time, illustrative composition and onboard/returned status. A record is returned only after its place in the stored-data queue has actually been transmitted; earlier downlinks do not count. The final debrief compares the sites and records elapsed time, flight time, battery energy used (net draw plus preheat) and peak battery temperature. It remains available after reloading. A reload resumes the last landed checkpoint and stops any in-progress sample or downlink; resume these explicitly.

Free flight and Sample here remain available and can satisfy expedition objectives. Samples collected before beginning do not count toward a new expedition. The original short survey is retained under Optional quick survey / Start over.

The landscape is an illustrative interpretation of the Ahmakiq Undae landing-area description: long dunes, broad low interdunes and distant hills, not the previous mountain basin. Dune spacing (~3.2 km) and heights (75-120 m) follow the published-size terrain retained during the merge; orientation and site positions are artistic training choices, not a surveyed elevation model. The planning map shades the same terrain, with lighter areas representing higher ground. Both the geology and science results are examples, not actual observations from Titan.

Quick Survey and Diagnostics

Surface operations queue is expandable beneath Science Payload in both views. Queue up to 12 fresh-sample, downlink or recharge steps, reorder/remove pending steps, then Run. The queue stays at its starting landing site. Each sample includes acquisition and analysis; downlink sends the data present when its step begins; recharge hibernates to the chosen battery percentage using the existing power/thermal model. 1x/20x/100x use the shared clock.

Unsafe conditions, manual takeover or interrupted work stop the queue for explicit review and Resume. Global Pause freezes the existing job; Stop queue cancels unfinished sampling without refund. Reloads preserve the plan/completed steps but never start it automatically. This is training automation, not official flight software. Extended samples still use estimated timelines. Clear the queue to plan operations at a different landing site.

Science Payload retains the 30-second demo and offers LDMS training (10 min) and GCMS training (30 min). Extended profiles have distinct stages, instrument loads and heat coupling to the shared bay; all timings and budgets are estimates, not flight procedures. Their instrument energy is approximately 25.7/129.8 Wh, plus vehicle and radio loads. Run them at 1x/20x/100x, pause or cancel. Unsafe conditions interrupt the job; completed records include their mode, site, budget and actual simulated downlink progress. Reloads stop unfinished jobs without awarding a result.

Methods, budgets & sources, Sample records, Science notebook and debrief and Engineering references are expandable. Section links jump to Flight, Objectives, Science, Notebook and References without removing any page content. See the science-profile source review and limitations.

Begin the interdune survey, fly to the dry outcrop manually or with guided flight, land, collect a sample and return to base. The outcrop sits at the edge of a rain-darkened interdune: damp ground (darker, safe to land on) around a small methane puddle (no landing). This matches the kind of ground expected near Dragonfly's equatorial landing area after a methane storm; it is training geography, not a reconstruction of the real site.

Diagnostics shares the live vehicle state: equipment/battery temperatures, circulation and insulation integrity, cold-duct trim, electrical load, MMRTG generation, battery reserve and approximate Titan local time. Hibernation on dry ground advances thermal/energy calculations with bounded one-second steps and stops accelerated time on warnings. Generation continues through day and night.

Read the source audit and model assumptions for the NASA, APL, Lockheed Martin and TubeTech references, including claims not supported by the public sources. The shell-and-tube figures illustrate the liquid-to-liquid exchanger concept explored in the design study; they are representative industrial designs, not Dragonfly flight hardware.

The independent liquid-to-liquid exchanger study calculates heat flow and outlet temperatures from explicit assumptions. It is not part of the flight hardware. See the research compendium and the release reconciliation for follow-up evidence.

Titan-inspired Weather

New missions get occasional gradual gust events and rarer methane-rain scenarios. Diagnostics offers Occasional weather events or Fixed wind / events off. Moving the surface-wind slider also selects Fixed wind. Automatic events stay within the 1.6 m/s design maximum; each starts with a visible 60-simulated-second advisory before building, peaking and recovering. Quiet gaps are 10-20 simulated minutes, with a 12% rain choice per event. These frequencies and times are training assumptions, not predictions for Titan. Pause freezes weather, and accelerated hibernation stops at a new advisory instead of silently skipping it.

Run strong methane-storm training is optional: 120 seconds of warning, a gradual build, and gusts up to 4.5 m/s above the flight design envelope. It holds new flight plan uplinks through recovery, but does not crash the aircraft or automatically abort an existing flight. Land before strong winds. Select Fixed wind to end an event; the current wind becomes the fixed setting until you change the slider.

Rain/cloud scenarios gradually thicken the Pilot-view haze, dim the light and add darkening/gloss to the existing rain-darkened interdune. Additional wetting fades with a two-hour simulated decay time. These are visual training effects: no flooding, puddle growth, new landing hazards or altered sample chemistry. Wind feeds the existing convection and power calculations, so component temperatures can respond; ambient remains -179.15 C, without invented Earth-style hot/cold fronts. Last-landed saves preserve event progress, randomness and wetness. Older saves retain fixed wind.

The inspiration is published Cassini observations of cloud evolution over 11 hours, equatorial methane-rain wetting, and inferred organic dust storms. They do not establish these local wind profiles or warning lead times.

Flight Planning

Real Dragonfly flights are planned on Earth, uplinked and flown autonomously (signals take 73-90 minutes each way). Plan flight opens a map: tap waypoints, choose 40 m, 150 m or 400 m cruise altitude and check the GO / NO-GO estimate (energy, time, battery heat, daylight, landing site). Uplink plan sends it; after a short simulated delay the lander flies itself. Use 1x / 5x / 20x to speed up time during the flight, or Stop plan to hold position.

Leapfrog scouting: the last waypoint must be a scouted (green) site. Flying within 60 m of an amber candidate site at 20 m or higher scouts it, so a later flight may land there.

Saved Progress

Your mission is saved in the browser while the lander is on the ground and whenever you leave the page: position, battery and temperatures, the Titan clock, scouted sites, samples, the science log, expedition notebook/debrief and stored data. Coming back resumes the last landed checkpoint. Start over (under Optional quick survey, or in Diagnostics) clears it and starts again from the landing sequence. Existing saves remain compatible; their expedition starts unbegun and their earlier samples remain in the Science Payload panel.

Vehicle Layers

Under the Mission diagram, Exterior / Internal / Thermal switches the vehicle view. Internal makes the foam shell see-through. Parts already redrawn from their own photos and papers: the sample carousel, Wonderwall and DraMS in the attic, the Frontier radio, the integrated electronics module, the two power switching units, and the lower DragonCam cameras (Microscopic Camera, LED array and Down Cameras in foam cavities), and the nose cameras (Forward MAC and WAC, two Side Cameras, two NavCams on a camera panel), and the DrACO drills (gold head with percussion and rotation actuators on a Z-stage rail; the head slides down while only the bit turns), and the sample-transfer system (flexible hoses, tubes into the attic, a flow diverter on the carousel and a return line to two blowers behind the port forward arm), and the 100 W radio amplifier (TWTA: a tube and its power conditioner under the top deck, drawn like published flight TWTAs because Dragonfly's is not pictured), and the high-gain antenna's gimbal (azimuth turntable and elevation drum under a foam dome just aft of the attic, the motorized arm and pedestal that raise the dish, and the launch and flight locks), and the battery (24 upright cells in a ribbed enclosure with an aluminum upper housing, fans and heat pipes, its look from real photos of APL's test module); their positions are mostly estimated. The rotor drive electronics are a provisional enclosure illustration (Codex) until a clear photo turns up. The aft bulkhead and EFIELD assemblies are also rebuilt; METHAN and Ocellus are approved design estimates, not photo-matched. Remaining placeholders are the battery/MMRTG temperature sensors; the nose bulkhead remains a provisional structural illustration, not a completed photo-match. The air loop was redrawn from NASA/APL's thermal talks: two circulation fans at the MMRTG end, the warm duct forward under the floor into the nose, and the trim device, a cold bypass duct under the belly between fore and aft diverter flaps (an earlier side-chimney concept is archived). Arrows follow the warm-air loop from the MMRTG. The loop is live: its arrows move with the fans' flow and go from MMRTG-warm to bay-cool; when the diverter flaps open, a share of the flow (blue) runs forward through the exposed cold duct, and the flaps swing open with the setting. Green dots mark the controller's two temperature inputs (battery and MMRTG fin root). A "Warm-air loop" panel under the diagram explains, with live values, how the air moves, how it is regulated (trim 0-40% in 2% steps about every 10 minutes) and how it is measured. Thermal colors every part by its live temperature. In either view, expand Component temperatures to see the live readings and their sources; it starts collapsed and is hidden in Exterior. Tap a diagram number to open the panel and select that part, or select its row by mouse, touch or keyboard. Both views color the readings with the same temperature scale, shown inside the panel; Scale zooms between whole-lander and inside ranges without changing the color of any reading or part. Colors are anchored to absolute temperature: blue is cold, green is the 0-20 C equipment/battery reference band, yellow is warming, and red is 35 C or higher. These are not fault indicators or certified operating bands for every component: the cold exterior and hot MMRTG gas have different targets. The Internal diagram keeps its component materials; Thermal colors the diagram by temperature. The Pilot view shows the same live battery, temperature, flight-time, link and plan readouts in a compact strip.

With the original model selected, Internal and Thermal show a clearly labeled mock-up interior instead: that model is not based on the real design, so its parts are made up for illustration (their temperatures still come from the simulator).

Each layer has its own backdrop. Exterior offers three, switched with the small a / b / c buttons at the bottom left of the diagram: (a) a cinematic dune panorama cropped from NASA/APL's Dragonfly animation, preserving the original terrain, lighting and haze; (b) an illustrative assembly clean room with a gently projected perforated floor, rectangular light reflections, keep-out tape, purge and tool carts, and larger equipment where space permits; (c) a composite of real Cassini Saturn and Titan photographs over a restrained star field. Both planets fit within the canvas, with the vehicle centered between them. A is an artist's concept, and C is not to scale or a view from Titan's hazy surface. Image credits are linked below the diagram; source and processing details are in the asset notes. Internal is an engineering blueprint with dimension lines for the published 3.85 x 3.85 x 1.75 m envelope and a title block; b turns it into a black sheet with a yellow grid and yellow dimensions, title and readouts. Thermal offers (a) Titan's -179 C air in the thermal scale's own color, dimmed so parts at air temperature stay visible, with a white grid, (b) light grey with a dark grid, like NASA's published thermal-model figures (the default: the cold blue parts stand out), and (c) black with a yellow grid. The same a / b / c buttons switch the backdrop in every layer, and each layer remembers its choice. Readouts turn dark on light backdrops and yellow in the yellow modes.

Science Instruments

The Science Payload panel shows each instrument working: DragonCam images in flight, DraGMet logs weather and occasional quakes, DraGNS measures the ground's makeup while landed (more precisely the longer it counts), and Sample here drills and analyzes the ground with DrACO and DraMS. Everything they collect is stored on board until a downlink sends it home; Comms shows how long that will take. Results are illustrative examples, not mission data.

Operations

• Motor preheat: the first liftoff after 30 minutes on the ground charges 60 Wh (8 motors x 90 W x 5 min, time-compressed).
• Comms: the Direct to Earth link is available only when landed, awake and in daylight. The antenna is stowed in flight. "Start downlink" draws an assumed 200 W and counts the data returned; it stops at takeoff, at night, in hibernation or below 30% battery.
• Flight endurance: Diagnostics shows the minutes left before the 15% reserve or the 35 C battery limit, plus a 30-minute flight timer. The alert bar warns "Land now" near either limit, after 30 minutes, or when flying at night.
• Wind above the 1.6 m/s design maximum is labeled a stress test.
• Titan day cycle (time-scaled): the Mission view's Titan day strip shows the day number, local Titan time, a day/night bar and the day's checklist in the real order of operations: send data home, one flight, sample science, then night (recharge, seismic and weather work). Sleep until morning fast-forwards the rest of the day and the night in a few seconds, wakes the lander at 08:00 local (after sunrise, so the scene is lit) and posts a morning report. A real Titan day lasts 382.7 hours (about 16 Earth days) with about one flight per day. Plans, sampling and downlinks can accelerate the shared simulation clock; rest is time-compressed too. A second flight in one Titan day is allowed but flagged.

Engineering Context

Mission and Pilot share the same vehicle mesh, flight state, eight rotor speeds, controls and local track. Pause freezes the simulation in both views. The terrain and slope-aligned rocks sample the same rendered triangles. Far from the vehicle, where the ground mesh is coarse, the terrain drops detail the mesh cannot show and averages dune shapes around each mesh cell, so distant ridges stay natural instead of saw-toothed; the ground the vehicle can reach keeps full detail. When the vehicle flies into a new 400 m square, the recentered ground is built a few rows per frame and swapped in when complete.

The APL thermal test report provides the Titan environmental baseline and nitrogen-chamber reference. The chamber is a thermal facility on Earth, not a Titan-gravity flight chamber. APL's rotor testing report covers NASA Langley and Sikorsky's aerodynamic work. Lockheed Martin provides cruise-stage and aeroshell hardware.

NASA publishes the 875 kg mass and 1.35 m rotor diameter baseline. Storage follows the 11.5 kWh design reported in 2022, not a newly verified flight-pack measurement. Estimated arrival output is 90 W (selectable 70 W), declining 2.5% per Earth year. The 0.75 figure of merit, drag area and 1.15 induced loss factor are assumptions. Coaxial pairs share four unique swept disks for the ideal momentum calculation. RPM mixing and the near-surface atmosphere approximation are educational, not flight-qualified. There is no chamber CFD, validated aerodynamic model or actual mission telemetry. The surface thermal model uses source-informed gas flow, foam and convection; its capacities, loss paths and PI gains are illustrative. Battery limits use the 2020/2023 designs; protective interlocks remain training logic. The trim closes in flight and uses 2% commands up to 40% on a 600-second landed control cycle. Both views use the three-blade design confirmed by the rotor engineer in 2026. Terrain-following altitude is a demo AGL coordinate, not an inertial vertical-dynamics solution.