hraness

saved

Defining AI Psychosis. Part 2: "Prolific AI Psychosis"

by Jeff Clark, MDJeff's Blogpublished

Hraness cites a source capture. The source author remains the source.

gist

Psychiatrist Jeff Clark defines prolific AI psychosis as high-volume AI output that fails to increase—or even destroys—real work value, driven by a soft break in quality judgment rather than by more code itself. State-of-the-art agent harnesses amplify the risk: intermittent slot-machine rewards, confident wrong decisions, and supervising toddler-plus-senior-engineer models make counterfeit wins easy to accept. The recovery he describes prioritizes human judgment, sleep, craft, and intentional boundaries over metrics of raw throughput.

ideas

  • Volume is not the disease. Prolific AI psychosis is failing to assess one's own output quality, so more lines of code or text can mask flat or negative value.
  • Harnesses raise the stakes. Agent loops that open programs, browse, and ship apps compound both miraculous wins and impossibly bad decisions that look like wins.
  • Supervision is the real job. Models speak with equal confidence as senior engineer and as red-Kool-Aid toddler; rejecting counterfeit wins takes skill, focus, and patience.
  • Intermittent reinforcement drives the spiral. Unpredictable rewards plus hype, job fear, and metrics-driven environments pull people into sleepless hyperfocus where usefulness questions feel like distractions.
  • Craft stays a human conversation. Useful, good, desirable, and delightful are hard for LLMs to supply; intentional boundaries matter anywhere AI can automate.

quotes

Prolific AI psychosis occurs when a person generates a large quantity of AI output without significantly increasing the real value of their work.

Jeff Clark, MD, defining prolific AI psychosis.

The problem lies in the subject’s perception of their output: they can’t assess the quality of their own work.

Jeff Clark, MD, locating the defect in judgment.

They’re one part senior engineer and one part toddler-running-across-white-carpet-with-a-jug-of-red-Kool-Aid.

Jeff Clark, MD, describing current AI models.

AI software development feels like playing the world’s most favorable slot machine.

Jeff Clark, MD, naming the intermittent-reward trap.