The productivity-experience paradox is the name Annie Vella and Kelly Blincoe give, in The Impact of AI Coding Assistants on Software Engineering: A Longitudinal Study, to a pattern of sustained positive perceptions of productivity alongside degrading developer experience among software engineers using AI coding assistants. The authors suggest that it may reflect a decoupling of two things prior research treats as linked — developer experience and productivity — when AI is embedded in development workflows.
Usage
The pattern comes from two questionnaires administered six months apart. Perceived productivity was positive and stable: 84% of participants reported improvement at both time points, and 77% of the matched cohort gave identical ratings. Over the same period, the share of matched participants reporting worsened developer experience in at least one dimension of the DevEx framework nearly doubled, from 14% to 27%. Feedback loops improved significantly, while cognitive load and flow state declined (not significantly), and flow state was the predominant issue within the negative cohort, rising from 54% to 76% of it. Changes in developer experience did not correlate with changes in productivity.
The authors’ explanation, drawn from participants’ open-ended responses, is that AI assistance accelerates throughput while redistributing effort into supervision, correction and verification: each suggestion must be evaluated and each generation verified, so the cycle of prompting, reviewing and iterating builds interruption into the workflow itself. They connect this to their proposed Supervisory Engineering Work, and suggest that faster feedback may be compensating for increased cognitive friction. Whether that reconfiguration is sustainable, or whether eroding flow and cognitive load eventually undermine the gains — for example as burnout or turnover — they leave as an open question.
As practical implications, they suggest organisations monitor developer experience alongside productivity, since output metrics alone may mask experiential erosion, and that how developer experience is defined and measured may need to evolve.
The paradox is reported from one longitudinal survey of 95 matched participants using perception-based measures, and its authors caution that sampling only continuing users of AI coding assistants likely overestimates positive outcomes.