LLM on Today I Learned - Saugihttps://til.saugi.me/tags/llm/Recent content in LLM on Today I Learned - SaugiHugoen-usMon, 19 Oct 2026 22:05:45 +0700LLM Reflexes and Guardrailhttps://til.saugi.me/llm-reflexes-and-guardrail/Mon, 19 Oct 2026 22:05:45 +0700https://til.saugi.me/llm-reflexes-and-guardrail/<p>AI will try to do something dumb. Not because it&rsquo;s a malicious model, but it&rsquo;s a probabilistic model. You give an AI agent a shell and it will execute <code>rm -rf</code> on anything it suggest worth clean up, or <code>git push --force</code> and rewrite the entire commit history.</p>