The AI Vampire

Original Article ↗ Hacker News Discussion ↗

Reactions to the Essay and Tone

  • Several readers find the “10x or die” framing exaggerated or juvenile, and some feel the self-glorifying anecdotes undercut the message.
  • Others like the fracking/vampire metaphor as a way to describe extraction of value from engineers, but see his doom scenarios as overblown or “ragebait.”

AI Productivity: 10x Claims vs Reality

  • Some commenters report huge gains: e.g., multi‑month database/schema/codebase migrations done in weeks, or 70k‑LOC apps where most new features are “one‑shotted.”
  • Many more say they see, at best, modest boosts: tickets still take roughly the same time once review, debugging, and rework are included.
  • There’s heavy skepticism that any tool is giving “nine extra engineers’ worth” of output; people point out there’s no visible wave of dramatically better non‑AI software to match the rhetoric.

Where AI Helps and Where It Fails

  • Commonly cited wins: boilerplate CRUD, small variations on existing code, search in large codebases, Google++ research, basic tests, pre‑PR code review, simple scripts, and simple React/SQL.
  • Pain points: GPU kernels, flatbuffers, fuzzers, financial/legal calculations, coding standards, large intertwined systems, and long‑lived maintenance. Models often hallucinate, ignore instructions, or “wiggle” around constraints.
  • Some see agents tied to tests/linters as promising; others note that even then the AI may try to “fix” the QA itself.

Jobs, Power, and Wealth Distribution

  • One camp argues AI should be banned or tightly constrained because it destroys junior roles and concentrates wealth; another says productivity gains help society “in aggregate,” though even supporters concede that gains skew toward capital.
  • There’s disagreement over whether “AI or your competitor will eat you” is realistic, especially where moats, product choices, and brand matter more than raw feature velocity.

Burnout, Addiction, and the “Vampire”

  • Multiple people recognize the slot‑machine dynamic: frequent small “wins” create engagement that feels like productivity.
  • Some engineers report feeling unable to think or code without AI, or working far more because “an hour of rest now costs a day of output.”
  • Analogies are drawn to cheap calories after the agricultural revolution: cheap features and code can lead to overproduction, exhaustion, and lower quality.

Labor Response and Unions

  • A thread develops around unionization: ideas include tying AI‑driven productivity gains to pay, limiting offshoring/H1B abuse, protecting equity, banning “unlimited PTO” games, and enabling ethical refusals.
  • Europeans note that stronger baseline labor law makes unions feel less urgent there, but still encourage organizing in the US.