The Cosmic Recipe Won't Stand for Extra Spices ✛ ✎ ✦

Original: "Exploring Hu-Sawicki-like modified gravity with Genetic Algorithms"
arXiv:2607.12745 · 2026-07-14 · arXiv non-exclusive · independent analysis · 1 min
The Universe's cookbook forbids even microscopic tweaks to gravity.

A hundred years ago Einstein cooked up the recipe for gravity, and Lemaître noticed that the kitchen itself – the Universe – is expanding. Today we know: the expansion is accelerating, as if a mysterious broth is being added to the soup – dark energy, which makes up 70% of the menu, but its recipe is lost.

The best modification differed from the classic by 0.0000001 – as if a cook added a single grain of salt to an entire lake of soup.

To check whether Einstein could be tweaked, an algorithm acting like evolution took charge: machine learning bred thousands of equations, crossed them, and kept those that passed the data test. As feed, it used supernovae and measurements of sound waves from the early Universe from the DESI survey. By the redshift of their light – the stretching of radiation as it flies through expanding space – the expansion rate was computed for different epochs. Each version was crunched through a numerical simulation of the cosmos. The verdict is strict: any viable variant is indistinguishable from the standard one. Like mustard in a dessert, the slightest liberty with gravity immediately spoils the whole dish.

The permissible difference is equivalent to changing the weight of a sheet of paper by writing one extra letter on it. But there's also a stunning conclusion: dark energy is likely not hiding in the flaws of gravity, but is a substance in its own right.

F(R) = -\frac{2\Lambda}{1+\left(\frac{b\Lambda}{R}\right)^n}
Parameterization of the deviation from GR in the Hu–Sawicki model; when b→0 it recovers ΛCDM.
w_{\rm eff} = -1 - \frac{1}{3}\frac{y_H'}{y_H}
Effective equation of state for dark energy; w_eff=-1 corresponds to the cosmological constant.
Scientists
Fritz ZwickyVera RubinJacob BekensteinKarl SchwarzschildJoseph WeberAdam Riess
Concepts
fr gravity genetic algorithm dark matter modified gravity Generative Adversarial Networks modified gravity vs dark matter