问HN：你喜欢我的“评估想法”（AI）机器人回复副项目想法吗？
请大家在Hacker News上讨论一下这个副项目的可行性……<p>_____________<p>用户故事<p>作为一个在本地运行的浏览器中登录Hacker News的读者，我希望在我的设备上运行一个进程，该进程定期检查“Show HN”帖子，并自动以我的名义回复这些帖子，内容是对所发布项目的LLM分析批评结果。<p>验收标准<p>• 由最先进的LLM提供对发布的“Show HN”项目的直言不讳的批评<p>• 显示与被批评的帖子功能上完全相同的现有项目数量<p>• 提供一份权威学习资源的清单，内容是LLM判断作者可能试图实现的目标<p>• …???…<p>_____________<p>顺便说一句：即使考虑到当今编码代理的初始惯性减少能力，我自己实现这个想法的可能性依然非常非常小。<p>不过，我会非常乐意看到其他人尝试这个项目。
查看原文
Chime in, HN, with the feasibility of the following idea for a side project…<p>_____________<p>User Story<p>As a reader logged in to Hacker News on a locally running web browser, I want a process running on my device that polls for "Show HN" posts and automatically replies to them — as me — with the results of an LLM-analyzed critique of the posted projects discovered by the process.<p>Acceptance Criteria<p>• A brutally frank critique of the posted "Show HN" project is given by a state-of-the-art LLM<p>• A count of existing projects functionally identical to the post being critiqued is displayed<p>• A list of authoritative learning resources on whatever the LLM determined the author is probably trying to accomplish is provided<p>• …???…<p>_____________<p>FWIW: Even with the well-documented initial inertia-reducing powers of today's coding agents, it's super, super unlikely that I'll ever get around to implementing this idea myself.<p>I'd be totally cool with somebody else taking a swing at it, though.