Daily Productive Sharing 1280 - Context Engineering

One helpful tip per day:)

Phil Schmid believes that true AI programming is, fundamentally, context programming:

  1. He quotes Tobi Lutke: “The art of prompt engineering lies in providing enough context for the LLM to have a chance at solving the problem.” He agrees completely with Tobi’s view.
  2. With the rise of intelligent agents, the information loaded into their “limited working memory” has become more crucial than ever.
  3. He observes that the key factor determining whether an intelligent agent succeeds or fails is no longer the model itself, but the quality of the context you provide. Today, most agents fail not because the model is broken, but because the context is poorly constructed.
  4. In his view, the gap between cheap demos and truly “magical” agents lies in the quality of the context you provide.
  5. As he understands it, the main job of code is not to “figure out how to respond,” but to “gather the information the LLM needs to achieve the goal.”
  6. He believes the real magic doesn’t lie in smarter models or more sophisticated algorithms, but in whether you provide the right context for the right task. That’s why he sees context engineering becoming so important.`

If you enjoy today's sharing, why not subscribe

Need a superb CV, please try our CV Consultation


Phil Schmid 认为真正的 AI 编程应该是上下文编程:

  1. 他引用 Tobi Lutke 的说法:“提示工程的艺术,在于为任务提供足够的上下文,使 LLM 有可能成功解决问题。”他认为 Tobi 说得很对。
  2. 随着智能代理(Agents)的兴起,加载到「有限工作记忆」中的信息,变得比以往更加重要。
  3. 他观察到,决定一个智能代理成功还是失败的关键,不再是模型本身,而是你为它提供的上下文质量。如今,大多数代理失败,原因不是模型失灵,而是上下文失误。
  4. 在他看来,廉价的演示和“神奇”代理之间的差距,就在于你提供的上下文质量。
  5. 在他的理解中,代码的主要任务,不是去“思考如何回应”,而是去“收集 LLM 为达成目标所需要的信息”。
  6. 他相信,真正的魔力,不在于更聪明的模型或更巧妙的算法,而在于是否为正确的任务提供了正确的上下文。这也是他认为,为什么上下文工程会变得如此重要。

如果你喜欢的话,不妨直接订阅这份电子报 ⬇️