The Programmer Identity Crisis

Original Article ↗ Hacker News Discussion ↗

Em dashes & AI detection

  • A large subthread debates whether frequent em dash use suggests AI authorship.
  • Some argue it’s now a reasonable heuristic in casual web writing; others say em dashes were already common (autocorrect, word processors, books) and people are just noticing them post‑LLM.
  • Several note that judging text as “AI slop” purely from em dashes is lazy and rude, and that accusations are affecting how humans write (e.g., avoiding dashes).

Programming: craft vs problem‑solving job

  • Many commenters resonate with the essay’s “craft” view: deep understanding, tinkering with tools, joy in writing elegant code.
  • Others insist coding is merely a means to solve business problems and pay bills; “fetishizing” tools and code style is seen as misplaced.
  • A recurring analogy contrasts chefs who love knives vs chefs who care only about the food; disagreement is over which mindset programmers should emulate.

LLMs in day‑to‑day development

  • Enthusiasts: LLMs speed up boilerplate, debugging, research, and “menial plumbing,” and can even make programming fun for those who never enjoyed it. Some report big productivity gains, new solo SaaS ventures, or using AI as a first‑pass reviewer.
  • Skeptics: describe “AI slop” PRs—thousands of added lines, hallucinated APIs, unused functions—which shift the real work onto reviewers. Brandolini’s law is cited: refuting bad LLM output is costly.
  • Several recount cycles of initial excitement, then retreating to using LLMs only for small, well‑bounded tasks after seeing quality issues.

Responsibility, process, and management

  • Strong view that authors remain fully responsible for AI‑assisted code; using “Claude wrote that” as an excuse is seen as unprofessional and grounds for rejection or firing.
  • Others note that leadership sometimes chases AI metrics (lines of code, tool usage), enabling bad behavior and burning out conscientious reviewers.
  • Open source maintainers report simply ignoring obvious AI‑generated patches due to review cost.

Identity, history, and the future of programming

  • Older developers recall “cowboy coding” days, see current AI trends as one more step in long‑running automation (COBOL, SQL, compilers, visual tools, SaaS).
  • Some predict hand‑coding will become a niche like knitting in the age of looms; others think LLMs may plateau and coexist as just another tool.
  • Many note an emerging divide: those who see themselves as hackers/craftspeople vs those who see themselves as general problem‑solvers whose identity isn’t tied to typing code.