技术英语精读 · 句式篇
📖 The new rules of context engineering for Claude 5 generation models — Thariq Shihipar, Anthropic (2026-07-24)
✍️ 句式拆解
原文：
“We removed over 80% of Claude Code’s system prompt for models like Claude Opus 5 and Claude Fable 5 with no measurable loss on our coding evaluations.”
🔍 结构拆解：
-
主句：
We removed over 80% of Claude Code's system prompt- 主语
We— 作者代表 Anthropic，第一人称增强权威感 - 动词
removed— 直接有力的及物动词，不做作 - 宾语
80% of Claude Code's system prompt— 数字+专有名词，数据说服力
- 主语
-
介词短语（对象限定）：
for models like Claude Opus 5 and Claude Fable 5for明确适用对象like表示举例而非穷举，留有余地
-
介词短语（结果/伴随）：
with no measurable loss on our coding evaluationswith引导伴随结果，句式紧凑no measurable loss— 否定+量化，精确有力on our coding evaluations— 限定测评范围，避免过度推断
💡 亮点：
- 数据驱动叙事：
80%具体数字给出冲击力，no measurable loss消除顾虑。好的技术写作用数字构建信任。 - 信息密度：一句话包含 who、what、scope、result 四层信息，无冗余词。
- 节奏控制：长句通过介词短语分段，读起来不喘——主句短促有力（7 词），后续短语逐层递进。
- 精确克制：
on our coding evaluations而非模糊的in our tests，限定范围避免过度宣称。
🎯 模仿练习：
We reduced our API response time by over 60% for high-traffic endpoints like search and recommendation with no degradation on accuracy benchmarks.
🧩 段落精读
原文：
“Overall, we found that we were overconstraining Claude Code, both through our system prompt and in our CLAUDE.md files and skills. For example, when we read transcripts of our own internal usage of Claude Code, we see several conflicting messages in a single request like ‘leave documentation as appropriate,’ or ‘DO NOT add comments’ as our system prompt, skills, and user requests clash with each other. Generally, Claude can interpret the user’s intent to get to the right answer, but Claude must think more carefully about these overlapping and conflicting messages before deciding what to do. And while these constraints were once needed to avoid worst case scenarios, we have since found we can delete many of them and let the model use surrounding context and judgement instead.”
🔗 逻辑结构：
- 总论点（Topic Sentence）：
we were overconstraining Claude Code— 直截了当的结论先行 - 具体例证（Evidence）：引用内部使用记录中的冲突指令
"leave documentation as appropriate"vs"DO NOT add comments"，用具体例子让读者直观感受问题 - 分析过渡（Analysis + Transition）：
Claude can interpret...but Claude must think more carefully— 先让步（模型能处理），再指出代价（需更多思考资源） - 结论（Resolution）：
delete many of them and let the model use surrounding context and judgement instead— 提出解决方案，回扣论点
📌 连接词分析：
| 连接词 | 功能 | 注释 |
|---|---|---|
Overall | 总起概括 | 标志段落核心论点即将出现 |
For example | 举例说明 | 从抽象论断转向具体场景 |
Generally | 泛化总结 | 把例子提升回一般规律 |
And while | 让步转折 | 承认历史合理性，引出新做法 |
instead | 对比替代 | 新旧做法对比，强化论点的进步感 |
💡 写作技巧：
- 结论先行 + 证据后置：先告诉读者重点，再解释为什么——这是专业技术写作的标准套路
- 阴阳句式：
Claude can...but Claude must...— 能力 vs 代价的对比制造张力 - 时态切换：
were once needed（过去）→have since found（现在完成）→ 强调时间维度的进步感
🎯 模仿练习：
Overall, we found that we were over-engineering our CI pipeline, both in build steps and in test configurations. For example, when we audited a typical PR, we saw three separate linting stages that checked overlapping rules as ESLint, Prettier, and our custom validator competed for attention. Generally, the pipeline still produced correct results, but each redundant check forced developers to wait an extra 2-3 minutes before merging. And while these layers were once needed to catch edge cases in older codebases, we have since consolidated them into a single unified check and let the linter configuration handle the rest instead.