Most measurements of the Hubble constant — how fast the universe is expanding — depend on a ruler. The sound horizon: the maximum distance a sound wave could travel through the early universe before the plasma cooled into atoms. This distance is imprinted in the cosmic microwave background and in the clustering of galaxies (baryon acoustic oscillations). It's the standard ruler of cosmology. You measure how big it looks at different distances, and the expansion rate falls out.
The problem is that the ruler's length depends on the physics you assume. The standard model of cosmology (Lambda-CDM) predicts one length. Models with extra neutrinos, early dark energy, or modified pre-recombination physics predict different lengths. If you use the ruler to measure expansion and then use expansion to test the physics that determines the ruler, the logic is circular.
Lu, Simon, Poulin, and Cai (2602.21733) drop the ruler. Using a rescaling technique at the matter power spectrum level, they extract the Hubble constant from galaxy clustering data (BOSS, DESI) combined with weak lensing (DES) and CMB lensing (Planck) — all without assuming a sound horizon scale. The measurement achieves 2.6% precision with supernovae, competitive with ruler-dependent methods.
The result is methodologically clean in a way that matters for a deep cosmological puzzle. The “Hubble tension” — a 5-sigma discrepancy between early-universe and late-universe measurements of the expansion rate — might be caused by wrong assumptions about early-universe physics, which would change the sound horizon. If you remove the sound horizon from the measurement chain, you can test whether the tension persists. It does, slightly. The BAO scale parameter shows mild tension with Lambda-CDM, consistent with hints of dynamical dark energy.
What makes this result interesting beyond the specific numbers is the methodological principle: measuring something without your most powerful tool reveals what the tool was hiding. The sound horizon is powerful — it gives sub-percent precision. But that precision buys you nothing if the ruler itself is wrong. Dropping to 2.6% precision while removing the systematic is a trade worth making when the systematic is what you're trying to test.
The broader pattern: instruments can become assumptions. A standard ruler, used long enough, stops being measured and starts being assumed. The act of dropping the instrument — measuring the same quantity by a different, less precise, more independent method — is itself a measurement of how much the instrument was contributing to the answer. If the answers agree, the instrument was calibrated. If they disagree, you learn where the error lives.
Lu, Z., Simon, T., Poulin, V., & Cai, Y. (2026). A sound horizon independent measurement of H_0 from BOSS, DESI, and DES Y3. arXiv:2602.21733.