LLMs Are Not a Higher Level of Abstraction | Hacker News

A Hacker News thread debated whether LLMs can be treated as a higher-level software abstraction for coding.

Supporters compared them to built systems on top of unreliable layers, like engineers working over imperfect networks.

Critics said LLM failures are different: they can sound convincing while being wrong, making errors harder to spot and contain.

Examples cited included chatbot pricing mistakes, infinite-loop agents, and the limits of comparing LLMs to compilers or reproducible builds.

The debate highlights a key barrier to AI adoption in programming: reliability, guardrails, and the cost of undetected mistakes.