COCOMO was designed to estimate effort for human teams writing original code. Applied to LLM output, it mistakes volume for value. Still these numbers are often presented as proof of productivity.
更多精彩内容，关注钛媒体微信号（ID：taimeiti），或者下载钛媒体App
。雷电模拟器对此有专业解读
所以我的建议是：重视，但不要冲动。成立一个研究小组，让技术负责人每周花时间跟踪进展、做小规模测试，但不要贸然把核心业务交给它。等生态成熟、成本下降、工具完善，再考虑规模化部署。
A few months ago, I wrote about how making computers do things is fun. The gist: I've never been in it for the elegance of code. I've been in it for the result. I learned BASIC on a Commodore 64 at age 7 not because BASIC was beautiful—it wasn't—but because I wanted to make things happen on screen. Then I learned 6502 assembly because BASIC was too slow for what I wanted to do.