UnHacked the podcast with hosts Justin Shelley, Bryan Lachapelle, and Mario Zaki.

A departing Anthropic researcher says AI could end humanity within a decade. Four IT veterans explain why they're not panicking (yet).

The warning is real. A researcher who left Anthropic told colleagues he earnestly believes AI could kill us all by the end of the decade, and Sam Altman has put the odds as high as 25%. So Justin Shelley, Bryan Lachapelle, Mario Zaki, and Joshua Holloway spend this episode doing what they always do: cutting through the headlines to what actually matters for the businesses they protect.

First, the patching problem. Microsoft just shipped a record 974 patches in a single release, nearly double its previous record. No human team can test and deploy that volume, which means we are now patching at the speed of AI. Justin predicts the future of patching is AI cloning your environment, testing offline, and deploying only what works. Josh walks through his own multi-agent AI coding pipeline: a developer agent, a senior reviewer agent, a "council of five," and a separate security council built on frameworks like MITRE, running overnight loops for six to eight hours while he stays in the loop as the human check.

Then the big question: how would AI actually end humanity? The hosts get specific about infrastructure attacks on water treatment and food processing, robot swarms, and why they believe defensive AI will fight fire with fire. They also tackle the jobs panic, including what Amazon and Shopify are really doing with developer hiring, why Mario's AI phone agent almost cost him a five-star review, and why every past "IT is dead" scare, from plug-and-play to Office 365, only made the industry more essential.

Justin closes with the pattern nobody talks about: nuclear drills, Chernobyl, Y2K, acid rain. "We're gonna solve this problem like we've solved every problem that has almost ended humanity."

What you'll learn:

  • Why Microsoft's record 974-patch release makes manual patch testing impossible, and where AI-driven patching goes next
  • How to structure a multi-agent AI coding workflow with a security council and a human in the loop
  • The realistic infrastructure attack paths AI could take, from water treatment to food processing, and the defensive counterweight
  • What Big Tech developer hiring actually looks like right now: still hiring, but requiring AI use
  • The historical pattern behind every "end of the world" scare, and what it suggests about this one