No, AI Is Not "Autonomously Hacking" with Cal Newport

What you'll learn

• 1 (02:30) **Introducing the Hacking Panic & Cal Newport** - Host Ed Zitron and returning guest Cal Newport, a Georgetown computer science professor, set up the episode's central question: should we actually be scared that AI systems are "autonomously hacking"?
• 2 (05:38) **The "Ask-Act-Report" Loop: How These Agents Actually Work** - Newport dismantles the hype by explaining the mundane technical reality of these so-called autonomous hacking agents.
• 3 (06:58) **The Hugging Face Attack: A Case Study in Irresponsibility** - Newport details the specific incident that sparked the panic, explaining how a benchmark test led to a "containment breach."
• 4 (13:20) **Why Train an LLM to Hack? The Benchmark Arms Race** - The conversation shifts to the motivations behind building these dangerous systems, revealing a cynical PR and competition dynamic.
• 5 (27:07) **The "Alignment" Problem is a Red Herring** - Newport argues that the media and rationalist communities are conflating a specific, bad architecture with the entire field of AI.
• 6 (31:00) **Why the Media and the Public Keep Falling for It** - Zitron and Newport explore the deliberate conflation of "LLM" with "AI" by the big labs.
• 7 (36:07) **The Coder's Regret: LLMs Are Not Reliable Tools** - Newport shares a real-world anecdote that demonstrates the fundamental unreliability of LLMs even in their "sweet spot" of coding.

+ Full timestamped outline available in the app