---
date: "2025-09-16T00:00:00Z"
categories:
  - linkedin
  - llms
description: "Coding agents can over-literalize instructions, so sloppy or conflicting guidance in rules files creates predictable but absurd failures."
keywords: ["Codex", "AI coding agents", "instruction following", "AGENTS.md", "prompt design", "failure modes"]
---

GPT-5 (Codex) follows instructions exactly as given. Usually a good thing, but sometimes, it this is what happens.

**AGENTS.md**: ALWAYS WRITE TESTS before coding.

**Codex**: Let me begin with the tests. (Spends 5 minutes writing tests.)

**Anand**: Stop! This is a proof of concept. We don't need tests!

---

**AGENTS.md**: Write tests before coding. Drop tests for proof-of-concepts.

**Codex**: (Proceeds to delete all existing tests.)

**Anand**: STOP! We need those tests!

---

**AGENTS.md**: For new code, or if tests exist, start by writing tests.

**Anand**: Do this task. ... BUT SKIP TESTS! This is a POC!

**Codex**: The user has explicitly asked to skip tests. But the guidelines require tests for new code. For now, I think I will skip tests.

**Anand**: (sigh!)

Reminds me of Chaplin's Feeding Machine.

<div class="video-embed"><iframe src="https://www.youtube.com/embed/mVUVPen3xNQ" title="YouTube video" loading="lazy" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture" allowfullscreen></iframe></div>

[LinkedIn](https://www.linkedin.com/posts/sanand0_charlie-chaplin-demonstrates-the-feeding-activity-7370987866577022976-d1MA)