AI promised efficiency. Instead, it's making us work harder

Original Article ↗ Hacker News Discussion ↗

Impact on developer productivity and PR quality

  • Many report more numerous, larger PRs with hundreds of changes that are harder to review and often not understood by their authors.
  • Reviewers see “IDK, Claude did that” as increasingly common and universally unacceptable; this is compared to blindly pasting from Stack Overflow or BBS/magazine code.
  • Some say AI boosts individual speed, but teamwork suffers: less discussion, less shared understanding, more detached, “vibe-coded” systems.
  • Others note that AI tends to multiply whatever culture already exists: teams that previously shipped sloppy code now ship more, faster; teams with strong tests/use of abstractions use AI more responsibly.

Responsibility, review, and cognitive load

  • Strong consensus that humans remain accountable: if you submit a PR, you should be able to explain every line, regardless of origin.
  • Reviewers describe AI-generated PRs as time sinks: verbose, inconsistent explanations, brittle tests, unreachable code, and misleading comments.
  • “Parallelism” is seen as oversold: while AI can juggle many contexts, humans cannot; managing multiple agents is cognitively exhausting and can slow work overall.

Effects on junior/mid devs and learning

  • Several commenters say juniors now skip the struggle/learning phase, submitting working-looking code they don’t understand, which destroys trust and development of real skill.
  • Some teams respond by banning or sharply limiting giant AI PRs, or requiring overviews/presentations to force understanding.

Where AI feels genuinely useful

  • Positive anecdotes:
    • Rapid creation of small scripts, one-off automations, refactors, and boilerplate-heavy tasks.
    • Faster onboarding to unknown tech stacks, documentation lookup, and drafting designs/diagrams.
  • Key pattern: use AI for well-bounded, simple tasks you can fully verify; abandon it quickly when it’s near its limits.

Management, incentives, and labor dynamics

  • Many argue AI efficiency gains mostly benefit employers: same salary, more output, plus layoffs and work reallocation so individuals end up with 1.2× tools and 2× workload.
  • Historical parallels are drawn to cotton gin, self-checkout, industrial revolutions: productivity gains rarely translate into shorter hours.
  • Some connect this to broader capitalism critiques: productivity increases are absorbed as the new baseline; without unions or outcome-based pay, workers don’t capture much of the upside.

AI hype, economics, and article criticism

  • Mixed views on AI’s macro impact: some see an “AI bubble” sucking in capital and power; others argue successful AI investments could later broaden funding.
  • Commenters highlight flaws in the study cited (tiny sample, no AI experience) and say the article overgeneralizes, mixing “AI makes us slower,” “AI makes us faster but fills time with more work,” and “AI causes cognitive debt” without resolving the contradictions.