利用 Claude 的智能
One of Anthropic’s co-founders, Chris Olah, says that generative AI systems like Claude are grown more than they are built. Researchers set the conditions to direct growth, but the exact structure or capabilities that emerge aren’t always predictable.
Anthropic 的联合创始人之一 Chris Olah，说，像 Claude 这样的生成式 AI 系统更多是被培养而非构建。研究人员设置条件来指导成长，但出现的精确结构或能力并不总是可预测的。
This creates a challenge for building with Claude: agent harnesses encode assumptions about what Claude can’t do on its own, but those assumptions grow stale as Claude gets more capable. Even lessons shared in articles like this deserve frequent revisiting.
这为使用 Claude 构建带来了挑战：agent harnesses encode assumptions，关于 Claude 自身不能做什么，但随着 Claude 变得更强大，这些假设会过时。即使像本文这样的文章中分享的经验教训也值得经常回顾。
In this article, we share three patterns that teams should use when building applications that keep pace with Claude’s evolving intelligence while balancing latency and cost: use what it already knows, ask what you can stop doing, and carefully set boundaries with the agent harness.
在本文中，我们分享了三个模式，团队在构建应用程序时应使用这些模式，以跟上 Claude 不断发展的智能，同时平衡延迟和成本：使用它已经知道的东西，问问你能停止做什么，并仔细为代理框架设置边界。
1. Use what Claude knows
1. 使用 Claude 已知的内容
We suggest building applications using tools that Claude understands well.
我们建议使用 Claude 熟悉的工具来构建应用程序。
In late 2024, Claude 3.5 Sonnet reached 49% on SWE-bench Verified—then state of the art—with only a bash tool and a text editor tool for viewing, creating, and editing files. Claude Code is grounded in these same tools. Bash wasn’t designed for building agents, but it's a tool that Claude knows how to use and gets better at using over time.
2024 年末，Claude 3.5 Sonnet 在 SWE-bench Verified 上达到了 49%—当时是state of the art—仅使用一个 bash tool 和一个用于查看、创建和编辑文件的 text editor tool。Claude Code 基于这些相同的工具。Bash 并非为构建 agents 而设计，但它是 Claude 知道 如何使用并且随着时间推移使用得越来越好的工具。
Scores on the SWE-bench Verified benchmark across Claude model versions highlight its evolution.
Claude 模型版本在 SWE-bench Verified 基准上的分数突显了其演进。
We've seen Claude compose these general tools into patterns that solve different problems. For ...