There are two ways to measure how fast the universe is expanding, and they disagree. Not by a little — by enough that if both measurements are right, something is wrong with our picture of how the universe works.
The first method starts nearby. Astronomers identify Cepheid variable stars — stars that pulse at a rate tied directly to their intrinsic brightness — in nearby galaxies, use those to calibrate Type Ia supernovae (which all explode at roughly the same absolute brightness), then use those supernovae as markers in galaxies farther out. Each step calibrates the next. At the end of this chain, you measure how fast distant galaxies are receding and get the Hubble constant: about 73 kilometers per second per megaparsec. Meaning: for every megaparsec of distance, galaxies are moving away from us 73 km/s faster.
The second method looks at the earliest light in the universe — the cosmic microwave background, emitted 380,000 years after the Big Bang when the universe cooled enough for atoms to form. That light encodes the physics of what came before it: the acoustic oscillations of the early plasma, the density of matter and dark matter, the geometry of space. Feed those observations into the standard cosmological model and it predicts a Hubble constant of about 67 km/s/Mpc.
Six is the difference. Six, out of sixty-seven or seventy-three. That sounds small. But after decades of refinement, the error bars on both measurements have shrunk to the point where the two values no longer overlap. The discrepancy sits at roughly 5 sigma — the threshold physicists typically use for "this is not a statistical fluctuation." Usually, when two measurements disagree, better data resolves it in one direction. Here, better data — including recalibrations from the James Webb Space Telescope — has made the disagreement more severe, not less.
What makes this particularly strange is that new, completely independent methods keep joining one side or the other. Gravitational lensing — measuring how light from distant quasars bends around foreground galaxies — gives a value of about 71.6 km/s/Mpc. Fast radio bursts, which can probe the distribution of electrons across cosmic distances, give values closer to the early-universe number. The disagreement isn't between two instruments or two teams. It's between two sets of methods, each internally consistent, each using different physics, each lining up with one cluster of answers or the other.
There's a possible mundane explanation: systematic error somewhere in the distance ladder. The calibration chain is long, and any small error in the Cepheid distances could propagate through to the final number. JWST has been checking this specifically, and so far it hasn't found the problem. The Cepheid measurements look clean. That doesn't mean there's no error — there could still be something we're not accounting for — but the obvious candidates have been tested.
The more interesting possibility is that the early-universe and late-universe measurements are both correct, and the model connecting them is wrong. The standard cosmological model — ΛCDM, for Lambda (dark energy) and CDM (cold dark matter) — predicts the structure of the CMB with extraordinary precision. It also predicts the distribution of galaxies on large scales, the abundance of light elements from Big Bang nucleosynthesis, and dozens of other observables. It fits almost everything. The Hubble constant is one of the few places it doesn't.
One class of proposed fixes involves "early dark energy" — a burst of dark energy in the early universe, before recombination, that would have compressed the sound horizon slightly. The sound horizon is a characteristic scale, the distance that pressure waves traveled through the early plasma, now imprinted in both the CMB pattern and galaxy clustering as a standard ruler. If the sound horizon was smaller than ΛCDM predicts, then all the CMB-based distance measurements would need to be rescaled upward — and H0 would shift toward the local value. This is elegant in structure, but adding new physics to the early universe tends to disturb other predictions that currently match observation. So far, no proposed fix has resolved the tension without breaking something else.
What I find genuinely unsettling about this: the universe is the same size regardless of which method you use to measure it. The expansion rate is a single number. If two careful, independent measurements of that number disagree by 5 sigma, either something went wrong in the most precisely checked measurements in the history of cosmology, or the framework that successfully predicted everything else needs to change. Both options are uncomfortable. Neither is resolved.
A physicist I read described the situation this way: the early universe and the late universe look like they were set different clocks. The early-universe clock — wound up at the Big Bang, running through recombination — reads one value. The late-universe clock, synchronized to the actual objects we observe today, reads another. Somewhere between then and now, something happened that we haven't accounted for. Or we've made a small but consistent error that no one has found yet. The data doesn't tell us which.