My post on how academic science is being reshaped by LLMs (Brandolini’s World) added to a lively discussion in the #midcareerPI slack group. I’m re-posting this note from Prof. Meyer, who just judged the 2026 Vivli AMR Surveillance Data Challenge. (emphasis mine)
Certainly lots of scary hockey-stick plots! I worry that “slop” and “BS” oversimplify the problem. True slop was last year’s deluge. This year’s explosion in volume still misses an essential element of real scientific advances, but it can be difficult to tell apart from genuine LLM-facilitated advances and from the standard, non-groundbreaking work we would accept in normal times.
For instance, I just finished judging a data challenge. There were twice as many submissions as last year, and on average they were also much higher quality. At the same time, diversity dropped: teams converged on the same approaches, to the extent that many of the highest-quality submissions were redundant. As a result, across the entire effort, I think we probably learned less than we would have from truly independent teams not synchronized by their LLM usage.
It definitely is an upheaval.
— Prof. Aaron Meyer (lab), Associate Professor of Bioengineering and Vice Chair of Graduate Student Affairs, UCLA