Divide the box by the particle count and the design choice becomes legible. Spreading 166 billion particles across a volume 815 million light-years wide [5] leaves a mean spacing of roughly 148,000 light-years between them [13]. That average describes a mostly empty universe, and inside a collapsed halo the same particles sit far closer together, which is the sense in which Tiziana Di Matteo, the run's principal investigator at Carnegie Mellon, claims coverage over seven orders of magnitude in mass with the full physics [8][14]. The average is still the number worth holding onto, because it is what the volume cost.
Volume is also what the JWST comparison needs. The observations prompting this work suggest massive black holes were far more common early on than existing theory anticipated [11], and an abundance claim is a count per unit volume. Testing one requires a box large enough to hold a fair sample of rare objects while still resolving the gas inside them, which is a harder requirement than either alone. The team calls ASTRID the largest cosmological hydrodynamic simulation [4], and for this particular question the size is a functional requirement.
The endpoint does separate work. Redshift is a clock read backwards, with high z meaning greater distance and earlier times [15], and ASTRID began at z = 99 and ran 13.5 billion years to z = 0 [2]. A simulation that stops while the universe is still young can be adjusted until its early black holes resemble the observed ones, without ever answering for the galaxies and black holes we can count nearby. One that finishes has to land on both ends of the record. Zhou and colleagues published the completed run in The Astrophysical Journal in March 2026 [3].
What remains open is which channel actually produced those early black holes. Black hole feedback in a run this size is a prescription written by hand, and Zhou says the team is refining its gravitational and black hole feedback models as it prepares for new hardware [10]. A match between ASTRID's high-redshift output and JWST's counts would therefore be consistency between an observation and a model with adjustable interiors; the question of how those objects formed and grew so quickly stays open [11]. Di Matteo's own framing stops at making predictions and probing the puzzles [12]. What is new here is a full-history run at this volume that other groups can hold their own models against [1].
