2,095 words · 1 diagram
The high concept, the pillars every feature must serve, and an honest account of who this is for and what it is up against.
Read this and loops and you have the game. Everything else in this directory is depth.
| Version | Date | Author | Changes |
|---|---|---|---|
| 0.1 | 2026-08-19 | Jon Jaques | First edition, written against milestone 1 |
The one sentence
InertialRef is a first-person spaceflight simulator set in the real Milky Way, in which you can fly from interstellar space to a rock you can pick up without crossing a single mode boundary — and it runs in a browser tab.
The elevator pitch
Every star you can see from the cockpit is a star that exists. The nearby ones come from published astronomy; the rest are generated, deterministically, from a seed — and when new astronomy is published, the generated ones politely stand aside for the real ones. You fly a ship whose drive does not accelerate you, it reassigns which inertial frame you are at rest in, which is why you can cross a star system in minutes and still have to spend four of them shedding velocity before you arrive. There are no loading screens between orbit and the ground because there is no boundary there to load across. And the whole thing is a URL.
Why this is worth making
Three claims, in order of how load-bearing they are:
1. Nobody has built the continuous version. Elite Dangerous is seamless from orbit to surface but its planets are airless outside expansions and its galaxy is synthetic beyond the local bubble. Star Citizen is seamless and gorgeous inside a handful of hand-built systems and is not a galaxy at all. No Man's Sky is a galaxy and is not real, and its planet-to-space transition is a curated wipe. The intersection — real astronomy, continuous scale, one body, one cockpit — is empty.
2. The hard part is already done and it is provable. Milestone 1 was not a graphics demo; it was twelve executable claims about precision, determinism, addressing and frame-rate independence (the list, with measurements). One inch resolves to 9.4 µm at 8.18 kpc from the galactic center. Five hundred floating-origin rebases across 2,560 km produce zero drift. The same tick count produces the same state hash at 60 Hz, 144 Hz and 100× warp. Most projects with this ambition die on exactly these problems, years in, when it is a rewrite rather than a refactor.
3. Zero install is a distribution advantage nobody in this genre has. The comparable titles are 100 GB downloads gated behind a store account. This is a link. That is not a small thing for a genre whose central pleasure — look at this thing I found — is inherently social and inherently a thing you want to send someone. A portable photographic preset makes a composed view shareable by URL.
🎮 Designer's Note: The instinct will be to fight the browser and chase Star Citizen's fidelity. Resist it. The browser is not the compromise this project tolerates in order to ship — it is the reason the project is interesting. Fidelity is where we will always lose. Continuity, reality, and reach are where we can win outright. Every scoping argument in this bible resolves that way.
The four pillars
Every feature must serve at least one. A feature serving none is cut. A feature that violates one needs an ADR, not a discussion.
flowchart TB
P1["<b>One Continuous Space</b><br/>no mode you can name"]
P2["<b>The Sky Is Real</b><br/>truth first, procedure fills gaps"]
P3["<b>Momentum Is Law</b><br/>nothing stops instantly"]
P4["<b>You Are One Person</b><br/>one body, one viewpoint, always"]
style P1 fill:#0369a1,stroke:#0c4a6e,color:#fff
style P2 fill:#0e7490,stroke:#155e75,color:#fff
style P3 fill:#065f46,stroke:#064e3b,color:#fff
style P4 fill:#334155,stroke:#1e293b,color:#fffPillar 1 — One Continuous Space
There is no space mode and no planet mode. A player descending from orbit passes through half a dozen internal representation changes and must not be able to name a single one of them. No loading screen, no fade, no cutscene, no docking animation that hides a level swap, no "entering atmosphere" transition.
What this forbids: any feature whose implementation is easiest as a separate scene. Interiors are in the world. Stations are in the world. The galaxy map is a HUD overlay drawn over a still-running cockpit, not a screen you go to.
Already proven: frame transitions mid-flight, floating-origin rebasing, landing, and terrain geomorphing — a patch hands over to its parent with the geometry, the normals and the surface material arriving together. Still required: LOD hysteresis, cross-fade and sphere-derived impostors, which are what a body's own tier switch still lacks — see art.
Pillar 2 — The Sky Is Real
The catalog is truth. Procedure is what we do where truth is silent, and it defers when truth arrives. A star's spectral class, mass, and position come from published data wherever published data exists, and every derived property — luminosity, habitable zone, color on screen — follows from it physically rather than aesthetically.
What this forbids: inventing a nicer sky. If Barnard's Star is a dim red dwarf it is a dim red dwarf in the cockpit, and if that makes for a boring system then the boring system is the content. The interesting design work is making truth interesting, not replacing it.
The data is real; the image is photographed. The canopy is a sensor, not a window — an image composited from hull sensors with gain, integration time and a selectable response curve. That distinction is what lets the game be as beautiful as the cosmos actually is without falsifying a single number, because nearly everything that would make space more beautiful is already there and merely below the threshold of human vision. See art.
Consequence: the dataset changes underneath us, permanently and forever. That is not a bug to be managed but the source of a mechanic — see galaxy.
Pillar 3 — Momentum Is Law
Nothing in the game stops instantly. Not the ship, not the player, not a thrown object. Every travel mechanic — attitude thrusters, the main burn, even the jump — is fundamentally a problem of shedding velocity you already have, and the fun is in planning that well and then executing it.
What this forbids: a brake key. Arcade "space friction". A travel mode that snaps to a stop on arrival. Autopilot that makes the problem go away rather than solving it visibly.
How it is expressed: brachistochrone burns — accelerate, flip, decelerate. Not a throttle you correct, a plan you commit to.
Already proven: 6-DoF rigid-body integration, patched-conic gravity, drag, landing at free-fall accuracy within 0.03%.
Pillar 4 — You Are One Person
The player controls one person. First-person view places the camera at the character's head; optional third-person view follows the same embodied player. The ship camera can likewise follow the controlled hull. Camera choice does not add actors, fleet commands, or a strategic view that controls the world from above. The planetarium's orbit camera remains a passive way to observe that world.
What this forbids: fleet command, base-building from above, an RTS layer, an inventory screen that pauses the world.
Consequence: cockpit and suit instruments belong to the person using them. Where a real instrument could display a reading, the design gives it a physical home. Third-person controls add spatial awareness while movement, collision, and permission remain attached to the same canonical player.
The implemented camera and control boundary is recorded in ADR-0047.
Genre and positioning
Primary genre: first-person spaceflight simulator. Secondary: exploration/survey sim, with first-person interaction and light systems-RPG progression. Tertiary, later: 6-DoF ship combat; scarce, lethal on-foot encounters.
Positioned in one line each:
| What it is | What we take | What we do differently | |
|---|---|---|---|
| Elite Dangerous | The exploration loop, perfected | Fuel-gated route planning, first-discovery credit, cockpit discipline, the A–E module grades | A real local galaxy rather than a synthetic one; seamless atmospheric worlds as the base case; The Expanse-style burns instead of supercruise; no grind economy |
| Star Citizen | Fidelity and continuity in a few systems | Seamless ship interiors, physicalized interaction, first-person everything | Galaxy scale instead of set scale; ships in a browser tab instead of a 120 GB install; shipping instead of not |
| No Man's Sky | Procedural breadth and the joy of naming things | Discovery-as-reward, generation as content, planetary variety | Real astronomy underneath, real orbital mechanics, real momentum |
The positioning sentence, for external use:
Elite Dangerous's exploration loop, run on real astronomy, seamless all the way down to your hands, in a browser tab.
See competitive for the full analysis.
Audience
Primary — the survey pilot. 25–50, plays Elite Dangerous, Kerbal Space Program, Microsoft Flight Simulator, Outer Wilds. Reads a Wikipedia article about a star after visiting it. Owns a HOTAS or wants an excuse to. Values knowing where they are more than winning. Will spend an hour flying somewhere quiet and call it a good evening.
Secondary — the technically curious. Developers, astronomy hobbyists, and
people who will open the console and type ir.help(). The open-source posture and
the harness are aimed squarely at them, and they are also the contributor pool —
see sustainability.
Explicitly not the target: players who want a match to start, a lobby, a score, or a session that resolves in ten minutes. Nothing in this design is optimized for them and attempts to serve them will damage pillars 3 and 4.
Session assumption: 45–120 minutes typical, and the design must tolerate a 30-minute session ending mid-flight. The save holds dynamic state and references, and restores an identical state hash, so quit anywhere is already true and should be treated as a hard requirement rather than a nicety.
Business posture
Open source, non-commercial, no monetization of any kind. No purchases, no cosmetics, no subscription, no ads, no engagement metrics.
This is a design constraint before it is a business one, and a favorable one: with no revenue to protect there is no reason to build a grind, a timer, or a scarcity that exists to be relieved. Every reward in this game can be given for the reason it should be given — because the player did something interesting. See sustainability for governance, licensing, and the real cost question, which is hosting the persistent universe.
At a glance
| Working title | InertialRef (engine and game share the name) *(Resolved: engine and game share the name. The |
| Reference Drive fiction makes it diegetic rather | |
| than technical, and for an open-source project the engine identity is an asset.)* | |
| Genre | First-person spaceflight simulator / exploration sim |
| Platform | Browser (WebGPU); desktop wrapper considered later, not designed for |
| Audience | 25–50, simulation-literate, hardcore-patient |
| Modes | Solo offline · Solo online · Persistent universe |
| Business model | Open source, non-commercial |
| Team | One person directing coding agents |
| Engine | Custom TypeScript simulation core; Three.js today, WebGPU planned |
| Current state | Milestone 1 complete — 12/12 capability checks pass in Node and Chrome, online and offline |
| Named MVP | The Explorer — see production |
The honest constraints
Stated here so no page below has to relitigate them.
| Constraint | Consequence for design |
|---|---|
| One person and coding agents | Every system must be generated or simulated, never authored by hand at volume. A feature requiring 200 hand-made assets is not a feature. |
| Browser, WebGPU, ~10 W of GPU budget on a laptop | Geometric fidelity is well below Star Citizen. art spends the budget on light transport and HDR output instead, which is where this subject matter actually lives. |
| Non-commercial | No revenue to fund servers, so modes must make solo the complete experience and the persistent universe an addition. |
| No content pipeline | Ships, stations and interiors must be procedurally assembled from parts, not modeled. See content. |
| Real data, forever changing | Address stability and generation purity are non-negotiable. See galaxy. |
🎮 Designer's Note: The largest risk in this document is not technical. It is that the vision is a fusion of three games that cost, between them, something north of a billion dollars and thirty years, and this is one person. The mitigation is not optimism. It is production, which names one shippable slice — The Explorer — and treats everything else as horizon. A bible that pretends the whole thing is one milestone is a bible that gets ignored by month four.
Related
- loops — the game in motion
- competitive — the three comparables, in detail
- production — the MVP and the milestones after it
docs/vision.md— the platform charter this serves