My business partner sent a 5K vibe-coded PR that he didn't even test

Original Article ↗ Hacker News Discussion ↗

A software founder describes a business partner submitting a 5,000-line, AI-generated, untested code change, triggering doubts about continuing a career in programming amid the rise of “vibe coding.” Commenters debate whether large language models are merely tools that still require solid engineering judgment or a catalyst for widespread skill rot, lower standards, and a flood of low-quality code that non-experts can no longer distinguish from well-crafted work. Many see a growing split between engineers who use AI cautiously within strong testing and design practices and those who lean on it wholesale, with significant implications for careers, code quality, and how software is built.

Context: AI‑Generated 5K LOC PR

  • OP’s business partner submitted a ~5k‑line AI‑generated PR, apparently barely read or tested.
  • OP feels like a “Last Samurai” hand‑coding while others “vibe code” with AI.
  • This triggers broader discussion about AI’s effect on software quality, careers, and motivation.

Responsibility, Testing, and “Vibe Coding”

  • Many argue the core problem is not AI itself but bad engineering habits: not testing, not reviewing, shipping unrun code.
  • Others counter that AI amplifies those bad habits by making it trivial to produce massive, plausible‑looking code that no one understands.
  • Some use AI to generate code but insist humans must still design, run, and especially review tests and behavior.
  • Concern that AI‑generated tests can also be bogus, giving a false sense of security.

Impact of AI on Software Careers

  • Split views:
    • Some say experienced developers with “taste” and pre‑AI intuition will remain valuable; AI is just a faster tool.
    • Others think handcrafting code is a “dead end,” fear skill‑rot, and worry that laypeople plus AI will outcompete seniors economically.
  • Anxiety about interviews emphasizing AI proficiency over fundamentals, and about shrinking dev jobs, especially for newcomers.
  • Several consider career changes (teaching, consulting, gardening, yoga, security, GIS) or sabbaticals.

Diverging Attitudes Toward AI Tools

  • Camp A: AI as powerful assistant; good engineers can produce more, faster, and still uphold standards.
  • Camp B: AI as “vibe coding” enabler; encourages laziness, degrades skills, and leads to a looming “software apocalypse.”
  • Some report teams where undisciplined AI use is mandated or culturally enforced; others work in places that ban AI for codegen.

Business / Partnership Dynamics

  • Several emphasize that the deeper issue is mismatch in values between founders: rigorous testing vs “work smarter, not harder.”
  • Analogy to a marriage: if cofounders don’t share beliefs about quality, pace, and use of AI, the company stalls or self‑sabotages.
  • OP’s partner’s increasing reliance on AI seems to have eroded respect for OP’s expertise.

Skill, Taste, and Long‑Term Code Quality

  • Repeated theme: “taste” (architecture, UX, readability) still matters; AI can’t ensure good design.
  • Counterpoint: many clients accept mediocre outputs; good taste can be economically penalized if it slows delivery.
  • Some foresee roles cleaning up AI‑generated messes; others doubt such “un‑fucker” work will be allowed to be done properly.

Broader Social & Educational Concerns

  • Comparisons to social media, gambling, junk food, and drugs: AI optimizes for short‑term convenience and dopamine at long‑term cost.
  • Worries about “skill‑rot” and cognitive atrophy from over‑reliance on AI, similar to physical inactivity from car culture.
  • CS educators describe confusion over how to teach in an AI world: basics vs “vibe coding” vs higher‑level SE only.