There is no honorable path forward in a publish-or-perish system that rewards output over understanding. The researcher who shared this dilemma has put words to a tension many in AI feel but few admit publicly: either you churn out papers with whatever tools accelerate the pipeline, or you develop genuine expertise and fall behind in the metrics that matter to your career. Two papers written largely without LLMs, still unpublished. The temptation to let Claude carry the third from start to finish is not laziness, it is rational adaptation to an environment that does not reward diligence.
We have seen where this road leads. Reviewers at top conferences are already drowning in submissions that lack original reasoning, as When AI Slop Floods Top-Tier Conference Reviews makes plain. When researchers outsource the intellectual heavy lifting to generative models, the work becomes indistinguishable from slop, fluent but hollow. The reviewer assigned three ICLR papers who called them all slop is not an outlier; they are the canary. The system is consuming itself, and the individual cost is worse. Every paper vibed with Claude is a paper whose author has opted out of developing the mental models, the intuition, the hard-won skepticism that distinguishes a skilled researcher from an efficient prompter.
The real cost is not the missed publication. It is the eroded capacity to think clearly when the tools are not available, or when the tools produce convincing nonsense. The researcher who leans on Claude for an entire paper will struggle to catch subtle errors in the output, because they never went through the messy process of figuring it out themselves. They will produce more results, faster, and each result will be slightly less trustworthy. Over enough iterations, the field accumulates findings that nobody fully understands, built on code nobody fully debugged, stitched together by models that hallucinate with confidence.
We should ask what happens next. If the competitive pressure is already this intense for a researcher with two papers still seeking publication, how long before the only viable strategy is to bypass skill-building entirely? The security industry is already confronting this question from a different angle: when everyone has access to powerful AI tools inside an organization, as explored in When Everyone Uses AI, Enterprise Security Faces an Unseen Gap, the unexpected uses multiply faster than oversight can catch up. In research, the unseen gap is personal. The thing you lose by taking the shortcut is your own judgment. That cannot be recovered by publishing a third paper, no matter how fast Claude helps you write it.