I Remain a Skeptic
Points and comments are a snapshot, not live.
LLMs remain unproven for non-trivial software, with thin evidence and poor output quality.
Joshua Barretto explains why he remains skeptical of LLMs for serious software development, citing 4 years of industry 'revolution' with little to show, no clear productivity gains despite $1.5 trillion invested, and LLM-generated PRs that are still rubbish. He argues code is input, not output-citing Peter Naur-and views the push as homogenizing labor to reduce bargaining power. He does not feel he is falling behind by not using these tools.
What commenters are saying
Commenters split into two camps: skeptics who agree output quality is poor, and proponents who report personal productivity gains from judicious use. One skeptic notes studies showing only ~10% gains, not the 10x CEOs claim. Proponents argue that learning to use LLMs well takes time and they excel at debugging and style-transfer tasks. Some suggest LLM-driven velocity creates organizational friction, not just faster output.