Stop Letting AI Write Tests You Won’t Run: Enforce SDLC Gates That Catch LLM Hal

<h2> The Asymmetry of Change: Why Your Tests Are Looking the Wrong Way </h2> <p>A passing build is often treated as a certificate of correctness. In reality, it's a narrow

<h3> Intro </h3> <p>There's a pitch behind every "AI writes your tests too" workflow: more coverage, less manual toil, a safety net that used to take a sprint now takes

<p>An AI coding agent implements a feature, generates tests, runs them, and opens a pull request with a green build.</p> <p>That looks like independent verification. It often is

<p>Last month I asked a senior candidate the same question I ask everyone. "Tell me about a flaky test you fixed and how you fixed it."</p> <p>She paused and said, "I just added a

<p>I read the phrase "spec-driven development" and immediately decided it was for people without taste. Six months later, Claude Code generated a discount system that applied

<p>AI coding agents are getting really good at writing tests.</p> <p>But I kept running into one uncomfortable problem:</p> <blockquote> <p>Passing tests do not always mean strong

<a href="https://news.ycombinator.com/item?id=50133713">Comments</a>

<p>I used to treat repository instructions as the finished product.</p> <p>Write a careful <code>AGENTS.md</code>. Explain the architecture. Ban <code>any</code>. Require named<br