How do we know the universe's numbers?
Cosmology turns light into constraints through calibration, models, and cross-checks. Six evidence chains show where measurement ends and inference begins.
Lukasz Szramuk · · updated · 8 min read
No telescope has a dial marked Q, dark matter, or nuclear efficiency. Instruments count photons, record arrival directions, separate energies, and compare brightness. The constants and cosmic ingredients appear later, after calibration, foreground removal, statistics, and a model connect those signals to physical causes.
That distance between signal and claim is not a weakness. It is where science becomes checkable. A trustworthy result leaves a trail: another instrument can inspect the same sky in a different band, another team can change the analysis, and a rival model can try to explain the pattern with fewer assumptions.
Observation is what the detector records. Evidence is the way that record changes the odds among models.
A measurement is a chain, not a picture
The six stations below combine four observation images with two explicitly schematic measurement diagrams. Each separates the instrument signal from colour, calibration or reconstruction, then names the supported inference and the conclusion that would go too far. The final step opens a deterministic Tuner universe with exactly one control changed.
Evidence chain 01–06
Follow a claim from detector to dial
An image or diagram is the beginning of an argument, not its conclusion. Open each chain to see what an instrument records, what analysis adds, what cosmologists infer, and where the Tuner deliberately stops.
- 01Instrument signal
- 02Data treatment
- 03Supported inference
- 04Does not prove
Evidence and model boundary reviewed 28 August 2026
01 · Q · primordial structure
Planck: weigh the first wrinkles
How can ancient light constrain a number that later becomes galaxies?
ESA and the Planck CollaborationObservation source ↗ - Instrument signal
- Detectors measured microwave intensity across the whole sky in several frequency bands. After foreground removal, tiny temperature differences remain in the cosmic microwave background.
- Data treatment
- The familiar oval is a false-colour sky projection. Its colours encode temperature differences, while an angular power spectrum summarizes how much mottling appears at each scale.
- Supported inference
- The pattern constrains the amplitude and scale distribution of primordial fluctuations. Q is a compact toy-model stand-in for that early lumpiness, not a value read from one coloured pixel.
- Does not prove
- The map does not photograph inflation or select one unique theory of the early universe. Inference depends on a cosmological model and cross-checks among frequencies and datasets.
One-variable experiment
Remove the seeds
Open a universe with only Q moved below the Tuner's structure-forming window; every companion dial stays at our value.
Low Q shrinks the toy's coupled dark-energy ceiling so far that our unchanged Λ halts assembly first. The result is ‘lonely,’ not a direct image of smooth primordial gas.
02 · Q + N · structure growth
Webb: inspect what the seeds became
What can one deep field say about billions of years of structure growth?
NASA, ESA, CSA, STScIObservation source ↗ - Instrument signal
- Webb measured near-infrared light from a selected patch around SMACS 0723. Galaxies fill the field, while a foreground cluster stretches some background galaxies into arcs.
- Data treatment
- Multiple filters are assigned visible colours and combined. Distance estimates, lens models, selection effects, and completeness corrections are separate analytical layers—not details contained in the JPEG.
- Supported inference
- Deep fields test whether early fluctuations, gravity, dark matter, and expansion can grow the observed galaxy population over cosmic time. They connect initial conditions to later structure rather than measuring Q alone.
- Does not prove
- This field is not a fair miniature of the whole sky, and a beautiful arc is not by itself a precision measurement of a constant. Survey statistics and model comparisons do that work.
One-variable experiment
Make structure grow too violently
Hold the other eight controls fixed and strengthen gravity. This isolates one driver that the real observation constrains only jointly with others.
After URL quantization the preset crosses the toy's extreme-collapse threshold, so dense structure seeds black holes before a normal long-lived stellar population can form.
03 · χ · gravitational scaffolding
Bullet Cluster: compare two mass tracers
How do we map matter that emits no light?
X-ray: NASA/CXC/SAO; near-infrared: NASA/ESA/CSA/STScI; processing: NASA/STScI/J. DePasqualeObservation source ↗ - Instrument signal
- X-ray telescopes locate hot ordinary gas, while optical and near-infrared images locate galaxies. Background-galaxy distortions provide an independent gravitational-lensing signal.
- Data treatment
- Pink marks X-ray gas and blue marks mass reconstructed from lensing. Both are derived layers registered over the same collision; neither colour is visible to a nearby observer.
- Supported inference
- Most lensing-derived mass is offset from most collisional gas. Within the broader cosmological evidence network, that separation strongly supports a gravitating component unlike ordinary plasma.
- Does not prove
- Blue is not a photograph of dark-matter particles, and one cluster does not determine χ on its own. Geometry, lens reconstruction, collision dynamics, and alternative-gravity models must all be tested.
One-variable experiment
Take away the dark scaffolding
Open the Tuner with only χ reduced. The preset is a causal thought experiment, not a replay of the Bullet Cluster collision.
Visible matter gathers later and more weakly because the toy removes most of the early gravitational wells.
04 · ε · stellar chemistry
Cassiopeia A: read the ashes of a star
How does an X-ray shell tell us that stars made heavy elements?
NASA/CXC/SAOObservation source ↗ - Instrument signal
- Chandra separates X-ray photons by energy and position across the supernova remnant. Different ionized elements produce characteristic spectral lines.
- Data treatment
- Published colours are assigned to selected energy ranges and element maps: they are a readable key, not the remnant's natural appearance. Spectral models turn counts into element distributions.
- Supported inference
- The mapped silicon, sulfur, calcium, and iron are evidence of nuclear processing and explosive dispersal. Together with stellar models and laboratory nuclear physics, remnants test how stars enrich later chemistry.
- Does not prove
- The plate does not reconstruct every reaction in the progenitor or directly measure the Tuner's ε. The toy compresses a network of nuclear rates into one directional control.
One-variable experiment
Break the first rung of fusion
Open a universe with only nuclear binding efficiency lowered. The preset tests the Tuner's declared abstraction, not a simulated Cassiopeia A.
Hydrogen fusion fails in the toy, so the stellar furnaces needed to make and disperse heavy elements never get started.
05 · Λ + Ω · expansion history
Type Ia supernovae: turn brightness into distance
How can an exploding star reveal that cosmic expansion changed pace?
Schematic after Riess et al. (1998) and Pantheon+ (Brout et al. 2022)Observation source ↗ - Instrument signal
- Repeated images measure a supernova's flux through time, while a spectrum identifies its Type Ia class and host-galaxy redshift. Nearby distance anchors calibrate the absolute scale.
- Data treatment
- Light-curve shape and colour standardize the peak luminosity; cross-survey calibration, dust, selection bias, peculiar velocities and covariance enter before brightness becomes a distance modulus.
- Supported inference
- Distance versus redshift is inconsistent with a simple matter-only, steadily decelerating history. The original high-redshift samples supported late-time acceleration; modern supernova compilations constrain it jointly with BAO and the CMB.
- Does not prove
- A supernova does not detect dark-energy particles or read Λ directly. The inference compares whole expansion models and remains sensitive to calibration, population evolution, dust, lensing, selection and the assumed geometry.
One-variable experiment
Let expansion win early
Open a universe with only the Tuner's Λ control raised. This is a directional contrast with our settings, not a fit to a supernova Hubble diagram.
The toy's expansion ceiling is crossed before galaxies assemble, producing isolated matter rather than the measured history of our universe.
06 · Ω + nuclear network · first nuclei
Primordial deuterium: audit the first three minutes
Why does fragile deuterium act as a density gauge for the hot early universe?
Schematic after Cooke et al. (2018) and Planck Collaboration (2020)Observation source ↗ - Instrument signal
- High-resolution quasar spectra resolve deuterium and hydrogen absorption in selected, very metal-poor gas clouds. Complementary low-metallicity emission spectra constrain primordial helium.
- Data treatment
- Analysts model velocity components, contamination and ionization, infer abundance ratios, then compare them with a Big Bang nucleosynthesis network whose predictions depend on baryon density, reaction rates and the expansion rate.
- Supported inference
- Deuterium falls strongly as baryon density rises, so its primordial abundance provides a sensitive baryon-density constraint. Agreement with the independently inferred CMB density is a cross-check of the hot Big Bang and nuclear calculations.
- Does not prove
- The abundance does not measure total Ω, the dark sector, or the Tuner's ε as a single knob. The toy merges total density and a nuclear-efficiency threshold where real BBN separates baryons, radiation, expansion and many reactions.
One-variable experiment
Thin out the cosmic inventory
Open a universe with only total density Ω reduced. The contrast tests the Tuner's later structure gate; it is explicitly not a nucleosynthesis calculation.
The toy allows early nuclei but dilutes matter before mature structure forms, exposing how its Ω control differs from a real baryon-density inference.
What the Tuner can teach
A controlled preset makes causality legible: keep eight controls at our values, move one, and watch which epoch disappears first. It is useful for asking directional questions such as whether weaker initial structure delays galaxies. It is not a fit to Planck, Webb, Chandra, or any survey. Its windows are pedagogical rules and its sky is an artistic mapping of the model verdict.
That boundary is why each experiment links back to a dial explainer. The real evidence usually constrains several parameters together, while the one-variable toy intentionally breaks those couplings apart. Learn from the contrast, but do not mistake a dramatic preset for a numerical prediction about another universe.
How a number earns confidence
Confidence grows when different records cross-check one another. The microwave background, element abundances, galaxy clustering, gravitational lensing, supernova distances, and expansion surveys overlap without being copies of the same experiment. Their disagreements matter too: a tension may expose underestimated uncertainty, an unmodeled systematic effect, or physics missing from the standard model.
So the honest answer to ‘how do we know?’ is neither ‘the image proves it’ nor ‘it is all model-dependent.’ Data and models meet in a chain whose links can be inspected. The strongest cosmological claims survive changes of instrument, wavelength, analysis, and assumption—and state clearly which links remain provisional.