Plausibly Wrong

AI writing its own tests is theatre

Sun, 19 Apr 2026 00:00:00 +0000

An org. chart slide is up on the screen in your AI strategy deck. Three boxes, clean arrows, comforting symmetry: “Code Agent” → “Testing Subagent” → “Review Agent.” Somebody says “defense in depth,” and the room nods because it looks like the software org. charts you grew up trusting.

But you’ve seen what happens next. The code ships with a bug that the tests “proved” couldn’t exist. The review agent writes a confident paragraph about why the tests are sufficient. And when your team digs in, you find the same blind spot echoed three times, dressed up as process.

Editorial standards

Mon, 01 Jan 0001 00:00:00 +0000

Plausibly Wrong is published by Zwischen. We write about how AI systems actually work, where they break, and what that means for the decisions you’re making.

How we publish

Every post is edited by a human before release.
We verify factual claims we consider material — numbers, quotes, product capabilities, legal and regulatory assertions.
We aim to distinguish clearly between observed behavior, interpretation, and recommendation.

Tools

We use modern tools — including AI — during research and drafting. Tools don’t get a byline. The editorial team is accountable for what we publish.