LLMs Are Not Fun

Sources of Fun in Programming

• Commenters split between:
  • Enjoying the process and craft: thinking through problems, typing code, understanding systems end-to-end, tight feedback loops.
  • Enjoying the result: shipped products, solved business problems, weird side projects that would never get built otherwise.
• For the first group, LLMs feel like “babysitting a robotic intern” and rob them of the satisfying parts (debugging, careful design, manual refactors).
• For the second group, LLMs are “intellectual crack” that remove drudgery and make previously impossible or too-costly projects feasible.

LLMs vs Autocomplete and Traditional Tools

• Some argue LLMs are just “autocomplete++”: another step in a long trend (IDEs, refactor tools, higher-level languages).
• Others insist they’re qualitatively different:
  • Generative, non-deterministic, and prone to hallucination.
  • They choose approaches and architectures, not just syntax completions.
• This leads to a new relationship category: not a passive compiler, not a teammate, but a confident stranger whose output must be audited.

Productivity, Code Quality, and Architecture

• Pro‑LLM experiences:
  • Dramatic speedups for CRUD apps, webshops, Home Assistant setups, internal tools, ops scripts.
  • Offloading boilerplate, repetitive refactors, test writing, API glue, and “yak-shaving”.
• Skeptical experiences:
  • High cognitive load from reviewing verbose or incorrect code.
  • LLMs struggle with architecture and domain modeling; seniors say the bottleneck is rarely typing.
  • Worry that “stochastic programming” produces systems no one truly understands.

Workplace Pressure and Job Security

• Several describe being effectively forced to use LLMs by management or peer expectations.
• Anxiety that if humans only do the “interesting parts” now, future models will eventually do those too, turning many developers into replaceable “boilerplate”.
• Others counter that tool adoption has always been uneven, that LLM productivity gains are overstated in many domains, and that organizing around work/wealth issues matters more than rejecting tools.

Tool Neutrality, Ownership, and Culture

• Disagreement on whether LLMs are “just tools”:
  • Critics note they mediate thinking and creativity, centralize power in a few companies, and may be weaponized against workers.
  • Supporters see them as like screens or tractors: context-dependent, with both good and bad uses.
• There’s recognition of strong emotional polarization:
  • Pressure in some circles to loudly love AI; in others, “AI bad” earns easy approval.
  • This post is seen as a “scissor statement” that cleanly divides people by what they value in programming.