Reasons to Dislike AI Coding
AI-assisted coding is provoking a split between those who see it as a powerful productivity tool and those who view it as a threat to craft, employment, and long-term software quality. Commenters debate whether LLMs should handle only boilerplate or increasingly “own” entire systems, raising concerns about cognitive atrophy, loss of junior roles, mediocre design shaped by model limitations, and codebases no one truly understands. Underneath the technical arguments lie broader worries about wealth concentration, dubious data use, and the stated ambition of AI firms to automate as many human jobs as possible.
Reactions to the Article
- Several readers find the piece unfocused and shallow, more like a list of talking points than an argument.
- The reliance on assertions like “we ought to seize the means of production” is seen as hand‑wavy.
- Others defend it as a personal blog post meant to externalize thoughts, not a comprehensive treatise.
Usefulness and Limits of AI Coding
- Many see LLMs as excellent for boilerplate, support logic, glue code, and higher‑level abstraction (modules, microservices), but not for critical core business logic.
- Some report large productivity gains and welcome the shift toward design and higher‑level thinking.
- Others argue LLMs often produce fragile, mediocre code that still requires heavy expert oversight.
Craft, Cognition, and “Code as Art”
- Strong theme: hand‑coding builds deep mental models, critical thinking, and pattern recognition; outsourcing too much risks cognitive atrophy.
- Analogies to calculators, GPS, washing machines, and CNC tools are debated; critics say these miss the long‑term maintenance and thinking aspects of code.
- Disagreement on whether code can be “art” or is just a tool that may produce art; demoscene is raised as a counterexample.
Employment, Power, and Political Economy
- Repeated concern that AI, like past automation, reduces headcount while concentrating wealth, leaving many worse off.
- Some argue employment levels follow overall economic activity, not specific tools; others cite falling entry‑level roles as warning signs.
- Discussion references Marx and “fully automated luxury communism,” with disagreement over how technology affects workers under capitalism.
Trust, Ethics, and Data Use
- Deep skepticism toward AI companies’ motives (profit, militarization, stock price), and doubts that they only aim to “help programmers.”
- Strong criticism that models are trained on work people did not intend for this use, discouraging open sharing.
- Concerns about copyright/“stolen” code contaminating projects and eroding junior‑developer pathways.
Tool Comparisons and Determinism
- Some liken LLMs to compilers or higher‑level languages; others reject this as invalid because compilers are deterministic, spec‑driven, and don’t charge per run.
Quality, Maintainability, and “Vibecoding”
- Complaints that LLM workflows encourage accepting “good enough” partial solutions and ignoring long‑term quality.
- Others claim modern models can produce cleaner code than many rushed humans, provided someone with domain knowledge is actively guiding and reviewing.