JWST has repeatedly shown massive galaxies shining brightly when the universe was only a few hundred million years old, creating a major puzzle because most theoretical models predicted far fewer such objects. A new preprint (arXiv:2604.05024, not yet peer-reviewed) tackles this from the opposite direction: instead of hunting for the earliest galaxies, it examines the recent star-formation histories of already 'dead' massive galaxies at later times (2 < z < 5) and asks whether their pasts are consistent with the observed rapid rise in quiescent galaxy numbers since z≈7.

The RUBIES survey team, led by Yunchong Zhang, combined JWST NIRCam broad-band photometry from CEERS and PRIMER with NIRSpec/PRISM spectra. They performed Bayesian spectro-photometric fitting using the Prospector code on 22 massive quiescent galaxies (M* > 10^10.5 M⊙), testing robustness against different stellar libraries (MIST, BPASS) and star-formation history priors. For each galaxy they derived the most recent ≈1 Gyr of its SFH, defined a 'quiescent visibility timescale,' and extrapolated the implied number density at earlier cosmic epochs. These reconstructed densities were then compared with direct observations, including a new measurement from the PANORAMIC pure-parallel survey that uses multiple independent sightlines to reduce cosmic-variance errors.

The match is striking: reconstructed and observed number densities agree within uncertainties up to z≈7. This internal consistency strongly supports the reliability of stellar-population synthesis at high redshift and implies that the progenitors of today's quiescent population must have formed stars in intense, short-lived bursts before rapidly shutting down. Previous popular coverage of JWST's 'impossibly early' galaxies (e.g., Labbe et al. 2023 in Nature, arXiv:2207.12446) emphasized the tension with simulations but largely missed this progenitor–descendant linkage; they treated each new high-z detection in isolation rather than tracing the population continuity through detailed SFHs.

Synthesizing the present work with both the Labbe et al. luminous-galaxy census and theoretical predictions from the IllustrisTNG and FLARES simulations (which under-predict quiescent number densities by ≳1 dex at 3 < z < 7) reveals a clear pattern: observations are now self-consistent across cosmic time, so the mismatch lies in the models. Current simulations appear to quench star formation too slowly or inefficiently at early times, possibly requiring stronger AGN feedback, different stellar initial mass functions, or burstier star-formation duty cycles than assumed.

Limitations must be kept in view. The spectroscopic sample remains modest (dozens of objects), typical for deep NIRSpec programs, and results are sensitive to SFH parametrization and potential contamination from obscured AGN. Cosmic variance is mitigated but not eliminated. As a preprint, the analysis awaits independent peer review.

By demonstrating that the star-formation histories of massive quiescent galaxies align with their rapid buildup since z≈7, the paper resolves a major tension raised by JWST’s early detections and tightens constraints on early-universe galaxy-evolution models. The narrative shifts from 'impossible galaxies' to 'inadequate simulations,' sharpening the next generation of theoretical work needed to explain how the first massive systems assembled so quickly.