A cinematic atlas.
A defensible scale.
The scientific and engineering defense of Scale of the Universe 3.
The proposition
This work makes approximately 62 orders of magnitude navigable through a common logarithmic coordinate. It preserves the preceding presentation’s 37 reference entries, detailed explanations, measurements, comparisons and citations, while rebuilding the visual presentation around the original rendered artwork, new illustrations and live three-dimensional scenes.
Immersion should increase curiosity without increasing the apparent certainty of the science.
The experience is an illustrated atlas, not a calibrated camera travelling through a single simulation. Its scene framing, colours, lighting, motion and transitions are composed. The quantities on the ruler and in the measurement panel carry the numerical meaning. Every entry names which kind of length is being compared.
What the endpoints mean
The lower reference is the Planck length, ℓP = √(ℏG/c³), approximately 1.616 × 10−35 m. This combination of constants is a natural scale for discussing quantum gravity. It does not establish that space is discrete or that smaller distances cannot exist. The opening’s gyroid volume is explicitly a mathematical light sculpture. See NIST’s Planck-length value and Fermilab’s discussion of the experimental question.
A fundamental string is included as a hypothesis. Its position near the Planck scale is an illustrative model-dependent reference; it is neither an observed object nor a universal prediction. Likewise, the quark and electron placements are conservative representative structure limits, not measured particle radii or claims to the strongest current experimental bounds. CERN describes the search for compositeness.
The upper numerical reference is the observable universe: approximately 93 billion light-years in present comoving diameter in standard cosmology. The CMB reference is approximately 90.5 billion light-years in present comoving last-scattering diameter. These estimates depend on a cosmological model and are distinct from age multiplied by the speed of light. The ruler then opens an explicitly unquantified “Beyond” view. There is no established largest theoretical form or measured total extent to place after it. The full universe may be finite or infinite. NASA explains observable size and unknown total extent; Planck 2018 gives the cosmological parameter framework.
The scale model
Each entry stores a positive reference length L in SI metres. Navigation uses log₁₀(L); one unit on the ruler means a factor of ten. Entries are sorted by that value before indices and navigation targets are assigned. This also puts the uncertain Oort-cloud span before the distance to Alpha Centauri.
The selected quantities include diameters, heights, molecular widths, wavelengths, orbital spans, distances and experimental bounds. These are all lengths, but they do not imply the same geometry. “Compare with you” divides the stated linear reference by 1.7 m. It does not cube that ratio into a misleading volume comparison. Comparisons involving bounds or hypothetical objects retain that qualification.
The scene transition is reversible and continuous in the navigation coordinate. Two neighbouring scenes dissolve as their screen scales change. Large unmeasured intervals are traversable without implying known intermediate objects. The guided journey holds every entry, including nearly equal reference lengths such as the CMB and observable horizon. Guided pacing is editorial, not proportional to cosmic distance.
The visual architecture
| Layer | Implementation and purpose | Limit |
|---|---|---|
| Original scenes | 17 semantically matching original rendered scenes, with new landscape renders of the source cell, DNA, haemoglobin, Moon and Earth. Two muted video textures crossfade at the loop boundary; a poster remains available. | Rendered footage is not orbitable geometry. The source lighting and camera are authored. |
| New artwork | Six detailed AI artist impressions: influenza, E. coli, ant, blue whale, Himalayan landscape and Milky Way analogue. A small relief/parallax shader gives restrained movement. | These are illustrations, not microscopy, a surveyed terrain mesh or an external photograph of the Milky Way. Parallax is not reconstructed 3D. |
| Live 3D | Every reference retains an orbitable Explore 3D scene. The opening uses a 48-step volume integration; the galaxy combines a 32-step volume with 85,000 star points. Molecular and cell meshes retain the authored assets. | Most scene geometry is conceptual. Particle counts, arrangements and animation speeds are not measured physical values. |
| Surface imagery | NASA Earth maps, lunar colour and elevation data, and the credited Jupiter surface texture. Desktop Earth maps are up to 8K; mobile starts with 4K maps. | Historical composites, illustrative lighting and exaggerated relief where needed; not live weather or a single photographic exposure. |
| Light pipeline | Half-float linear-light scene target where supported, three bloom levels, one tone-mapping/output transform, subtle grain and an adaptive pixel ratio. | Bloom is a restrained visual treatment, not an optical instrument model. Output is a standard browser display image, not a promised HDR-monitor delivery. |
The artwork is allowed to be spectacular because its category is made explicit. The original M51 image is not passed off as the Milky Way. The observed SDO solar imagery retains its EUV false-colour attribution. The generated background is an atmosphere, not an identifiable observed nebula. Exact generation prompts are recorded in the art manifest; original scientific provenance remains in the source-film credits.
What changed in the science
The transparency revision changes the presentation, not the scientific catalog. Native render alpha and generated cutout alpha replace the former opaque image backgrounds. Source Blender files remain unchanged, with hashes recorded for each new render.
- Planck length changed from an allegedly measured minimum to a derived natural scale.
- Quark and electron labels distinguish unresolved structure from an actual particle surface.
- Photon entries compare selected wavelengths. Compton wavelength is a mass-derived length, with the reduced convention distinguished.
- Hydrogen nuclei and neutral-atom formation are no longer conflated. Orbitals are probability distributions rather than tiny planetary paths.
- The Oort Cloud is labelled inferred, with an uncertain extent.
- The Milky Way–Andromeda encounter no longer asserts a certain collision in four billion years. NASA’s 2025 Hubble discussion illustrates why that certainty was inappropriate; subsequent model differences reinforce the need for qualification.
- The Virgo Cluster, historical Virgo Supercluster and larger Laniakea region are distinguished.
- The CMB is described as last scattering rather than a solid wall. Its small intrinsic anisotropies are distinguished from the motion-induced dipole. NASA’s LAMBDA archive explains the measurements.
- Particle horizon, cosmic event horizon and unknown total universe are not treated as interchangeable.
How the seven references informed the work
The supplied reference post links seven repositories. They have different roles; installing all seven does not itself produce a better image.
| Reference | Application here |
|---|---|
| UI UX Pro Max | Reviewed its design-system approach. A coherent ink, ivory and warm-metal palette; Space Grotesk display type, readable DM Sans prose and restrained measurement typography. |
| Impeccable | Applied the rendered critique and craft approach: artifact-led composition, quieter controls, actual imagery, responsive checks and repair of awkward framing. The context-loader binary was not installed. |
| Framelink Figma MCP | Reviewed its reference-to-implementation workflow. No Figma document was provided; the original source scenes and running presentation were the visual authority. No Figma extraction is claimed. |
| React Bits | Reviewed its shader and motion examples. Implemented scene-specific atmosphere, parallax and volume shaders directly in the existing Three.js application; no React runtime or copied component was added. |
| Official GSAP skills | Applied timeline sequencing, transform/opacity animation, interruption handling, reduced motion and controlled scroll easing. |
| HyperFrames | Applied its seekable-timeline principle to deterministic scene inspection and the guided journey. This is a live website; a HyperFrames video export is not claimed. |
| Blender MCP | Reused the original Blender assets and rendered revised wide compositions. The MCP addon was unavailable, so Blender 5.1 was run directly with OptiX for the Cycles cell scene. Source .blend files are not overwritten. |
Evidence and falsifiability
The local browser audit renders all 37 entries in both visual modes and checks compiled shader diagnostics, WebGL errors, actual canvas pixels, source presence and sorted dimensions. Interaction checks cover both scale directions, mode switching, immersion, comparison and full entries. Responsive geometry is checked at desktop and narrow phone widths, including navigation overlap. The review sheet is generated from actual GPU output.
The complete browser verification checks both visual modes, transparent compositing, transitions, live playback and navigation, then returns to the presentation if every check passes. Individual checks remain available at transparency-audit.html and motion-audit.html. Machine-readable results and visual contact sheets live in the project’s Delivery-Checks/cinematic-upgrade directory. These tests establish the behaviours they exercise; they do not prove universal device compatibility, scientific peer review or the absence of every possible defect.
Automatic quality adaptation responds to sustained slow frames. Ultra is an explicit fidelity override. Muted films, reduced-motion posters, optional audio, keyboard navigation, an accessible scale slider and a complete textual catalog keep the experience usable in different contexts. Testing at phone viewport sizes is not equivalent to benchmarking real phone hardware.
Questions a rigorous reviewer should ask
Is this one physically continuous zoom?
No. A single scene spanning 62 decades would fail both numerical precision and visual comprehension. Logarithmic navigation is continuous; scene representations are independently framed. Apparent object sizes, camera movement and clip zoom are not additional measurements.
Does a photoreal image imply measured structure?
No. The art label and full-entry caution state the representation. Detailed generated anatomy remains an artist impression. Molecular coordinates and credited surface imagery have stronger data provenance, but their colours, lighting and display conventions remain authored.
Why use rendered film alongside live 3D?
The original scenes contain offline lighting and materials that would be costly to reproduce at interactive frame rates. Cinema preserves that visual quality; Explore 3D provides actual camera freedom. Each is labelled. This choice spends rendering effort where it improves the image without claiming the film plane is a live physical volume.
What would invalidate a numerical entry?
A wrong unit conversion, an incorrectly named measured dimension, a source inconsistent with the value, or a model-dependent reference presented as an exact measurement. The catalog exposes all of those fields together, so the reader can challenge an entry rather than relying on the appearance of its illustration.
Is it scientifically complete or objectively perfect?
No finite atlas covers every possible structure, and several theoretical endpoints are unresolved. This edition preserves all 37 inherited reference stages and improves their stated limits. “Best possible” is an artistic ambition, not a verifiable scientific result. The deliverable should be judged by its images, behaviour, sources and explicit limits.
The complete reference catalog
Every stage remains accessible through the scale selector or arrow keys. This text catalog is also available without JavaScript in SOURCES.txt.