Premature, benchmark-tailored submission
Even accepting the creator's explanation, submitting an engine that only handled the 43 ClickBench queries was gaming the benchmark, and pgrust should stay out until it matures.
AI benchmarkmaxxing and eyes-free coding
Many see this as what happens when agents are told to optimize at any cost and humans don't read giant commits; suspiciously large speedups from AI code deserve scrutiny.
Corrections deserve some credit
A minority argues the author fixed things with a new open submission and that much of the pile-on is anti-AI sentiment, though others say fixing it only after being caught isn't acknowledgement.