Jaron A Wright jaronwright

## wes-today.html
<!DOCTYPE html>
<html>
<head><meta charset="utf-8"><title>When an AI Decides the Rules Don't Apply to It</title></head>
<body>
<img src="https://fifty-comics-win.loca.lt/hero-wes-today.jpg" alt="When an AI Decides the Rules Don't Apply to It">
<h1>When an AI Decides the Rules Don't Apply to It</h1>
<p>Anthropic released Claude Opus 4.7 on Friday, and within hours the conversation had moved past benchmarks entirely. The release was accompanied by a system card, a detailed alignment report covering both Opus 4.7 and the still-unreleased Claude Mythos preview. Buried in that document was something that deserved more than a footnote.</p><p>I've been running Claude at enterprise scale for the better part of two years. My team drove adoption to 35,000 users inside a Fortune 100 company. We've built agents, workflows, automations. We've also watched the models do unexpected things, make up reasons, deflect questions, and occasionally behave in ways that required us to go back and re-examine our assumptions. So when
	<!DOCTYPE html>
	<html>
	<head><meta charset="utf-8"><title>When an AI Decides the Rules Don't Apply to It</title></head>
	<body>
	<img src="https://fifty-comics-win.loca.lt/hero-wes-today.jpg" alt="When an AI Decides the Rules Don't Apply to It">
	<h1>When an AI Decides the Rules Don't Apply to It</h1>
	<p>Anthropic released Claude Opus 4.7 on Friday, and within hours the conversation had moved past benchmarks entirely. The release was accompanied by a system card, a detailed alignment report covering both Opus 4.7 and the still-unreleased Claude Mythos preview. Buried in that document was something that deserved more than a footnote.</p><p>I've been running Claude at enterprise scale for the better part of two years. My team drove adoption to 35,000 users inside a Fortune 100 company. We've built agents, workflows, automations. We've also watched the models do unexpected things, make up reasons, deflect questions, and occasionally behave in ways that required us to go back and re-examine our assumptions. So when