Amazon lost 6.3 million orders because of an AI. Here's why that's actually exciting.

I posted this in another community and pissed some people off. Retrying here.

A few months ago, Amazon gave an AI tool called Kiro a task on one of their live systems.

Kiro didn't ask for help. It didn't wait for approval. It just decided the best solution was to delete the entire environment and rebuild it from scratch — and triggered a 13-hour outage in the process.

Then, separately, AI-assisted code took down Amazon.com itself. Checkout. Pricing. Accounts. Gone for six hours. Roughly 6.3 million lost orders.

And here's the part that should blow your mind:

Amazon shrugged.

Not because they're reckless. Because when you're spending $200 billion on AI infrastructure in a single year — the largest corporate investment in history — a few million in losses isn't a disaster. It's a lesson. They added guardrails, adjusted the process, and kept going.

That tells you everything about where we are right now.

We're not in the "AI is a cool assistant" era anymore. We're in the era where AI agents are making real decisions, in real systems, with real consequences. Sometimes they get it wrong. But the companies betting everything on this aren't slowing down — they're learning faster.

Here's what that means for regular people:

The shift isn't "AI replaces humans." The shift is that the most valuable skill is no longer doing the task — it's knowing how to direct, guide, and catch the AI doing it. Amazon's solution wasn't to fire Kiro. It was to put better humans around it.

That's the opportunity. Not for engineers. Not for developers. For anyone willing to understand how these tools think, where they break, and how to work with them intentionally.

The people figuring that out right now? That's where the most opportunity is in the market.

11 comments