Developer criticizes blind trust in and over-reliance on AI
What happened
On October 1, 2026, a Hacker News front-page comment titled "Claude Says" criticized developers who substitute LLMs for their own judgment: running a prompt once and treating it as verification, which is really a false appeal to authority. The author's example: for a request to "build a secure internally hosted package," the AI's plan was to fork, build and maintain every dependency from source, rather than use the artifact repository's built-in on-demand patching. The author argues the AI boom lets bad vibe-coded plans pass without review, while changing one setting in an already deployed plan needs layers of approval.
Written by AI from the coverage · updated 2 hours ago
Coverage
Follow the reports to see the story from different sides.
- Hacker News front pageClaude Says
一篇题为「Claude Says」的评论批评开发者用 LLM 代替自身判断：把提示词跑一遍就当成验证，实为虚假的权威诉求。作者举例称，面对「构建安全内部托管包」的需求，AI 给出的方案是从源码 fork、构建并维护每一个依赖，而非使用制品仓库自带的按需打补丁功能。作者认为 AI 热潮让糟糕的 vibe coding 方案无需评审即可通过，而修改已部署方案的一个设置却要层层审批。
Heat over time
Not enough continuous observations to draw a trend yet.