https://www.chenk.top/en/tags/safety/Recent content in Safety on Chen Kai BlogHugoenWed, 22 Apr 2026 09:00:00 +0000https://www.chenk.top/en/claude-code-learn/05-hooks/Wed, 22 Apr 2026 09:00:00 +0000https://www.chenk.top/en/claude-code-learn/05-hooks/<p>If MCP is how Claude reaches out, hooks are how you reach in. They enforce the rules you care about, not just what you hope for.</p> <p><figure class="article-figure"> <img src="https://blog-pic-ck.oss-cn-beijing.aliyuncs.com/posts/en/claude-code-learn/05-hooks/illustration_1.png?x-oss-process=image/resize,w_1200/format,webp/quality,q_88" alt="Claude Code Hands-On (5): Hooks, or How to Stop Worrying About Yolo Mode — Chapter overview" loading="lazy" decoding="async" class="content-image"> </figure> </p> <hr> <h2 id="the-model" class="heading-anchor">The model<a href="#the-model" class="heading-link" aria-label="Permalink to this section" title="Copy link to this section">#</a> </h2><p>A hook is a shell command that Claude Code runs at specific moments. The two you&rsquo;ll use most are:</p>https://www.chenk.top/en/llm-engineering/11-safety/Mon, 06 Apr 2026 09:00:00 +0000https://www.chenk.top/en/llm-engineering/11-safety/<p>Safety has the worst signal-to-noise ratio of any topic in LLM engineering. There&rsquo;s a lot of philosophy, a lot of marketing, and not a lot of engineering specifics. This chapter is the engineering specifics: what RLHF actually optimizes when it talks about &ldquo;safety,&rdquo; how refusal calibration breaks, what red-teaming looks like in practice, the hallucination measures that actually predict customer impact, and the small but significant 2024-2026 papers (Sleeper Agents, refusal as a feature direction, weak-to-strong generalization) that should change how you think about alignment in production.</p>