Vibe Coding 项目为何总比预期慢 3-4 倍？求破局建议

Оригинальный заголовок: Trying to get over the hump on projects

Заголовок и краткое изложение на выбранном языке ожидают перевода.

Краткий обзор ИИ

一位缺乏编程经验的 Vibe Coding 爱好者反馈，自己用 AI 做出的几个复杂项目实际耗时都是预期的 3-4 倍。他通常先用 AI 规划再动手，几小时就能做出不错的原型，但随后加功能、修 bug 会陷入无数轮排查循环，一两天后每轮排查耗时几乎等于整个首次构建。他常用 Kilo Code 搭配 DeepSeek 或 GLM 模型，并因受挫而频繁更换编程助手，现寻求改进建议。

Полный текст

Полный текст на выбранном языке ожидает перевода. Пока показан оригинал.

As a vibecoding enthusiast with little actual coding skills, I've managed to build a few complex projects that I'm pleased with. But each one has taken 3-4x longer than I expected.

I usually scope a project out, develop a plan with AI and start building. Within a few hours, I have something pretty impressive. Then I start adding features > noticing bugs... going into countless troubleshooting loops.

By the time I'm a day or two into it, each round of troubleshooting seems to take as long as the entire first build of the app. I get frustrated with my coding assistant/LLM and start switching around to see if I have better luck.

I usually use Kilo Code with DeepSeek or GLM models, if that has anything to do with things.

Any useful pointers to make better progress?