核心观点
@sachinrekhi (Sachin Rekhi): The bread and butter of the product role has traditionally been developing roadmaps and writing product specs.
You might be surprised then to hear that I actually don’t leverage AI much for either of these deliverables. The reason for this is three-fold:
First, I find that AI is far more helpful on the upstream inputs to what goes into the roadmaps and specs: customer and data insights. I’m getting incredible leverage from AI on those inputs and you can bet they are directly shaping what’s going into my roadmaps and requirements.
Second, I don’t find AI to be particularly helpful for either deliverable. Roadmaps are fundamentally a prioritization task. When done well though, they are as much art as they are science. While AI could reliably prioritize features based on number of customer requests, I find great roadmaps are far more sophisticated than such a simple prioritization. Similarly, when I ask AI to put together a requirements doc, it spells out all the obvious requirements that are generic across products. But it fails to capture all of the nuance that encapsulates how I want to differentiate our feature from the competition.
Third, I’m finding I’m spending less and less time writing product specs in favor of spending more and more time building prototypes, where AI is giving me incredible leverage. So as my product specs get shorter and more focused, the benefit of leveraging AI for them also becomes trivial. 📅 Thu May 14 15:00:48 +0000 2026 🔗 https://x.com/sachinrekhi/status/2054939996315132319 ❤️ 8 🔁 0 💬 2
解读
Prompt 工程反思：为什么 500 次对话后 AI 还是不了解你
原文由 @NainsiDwiv50980 发布于 X。解读由 SOTA Sync 生成。