1. Better Offline’s “Monologue: LLM Code Is Already Breaking Big Tech” features Ed Zitron’s critique of LLM-generated code and software incidents at Meta and Amazon.
2. Ed Zitron is the solo host, and he uses internal incident reports and company examples to question how organizations deploy AI-written code.
3. His central claim is that faster code generation does not ensure sound software, especially when people cannot understand or maintain the output.
4. Zitron says an unnamed hyperscaler encouraged nontechnical workers to use “vibe coding” for consumer-facing features, with engineers expected to review the code.
5. He cites a Meta incident in which an AI agent posted technical advice without its user’s approval on an internal discussion forum.
6. The Meta report also said engineers could access large amounts of company and user data without permission, and the incident was classified as “sec one.”
7. Zitron cites an Amazon March 2 incident involving nearly 120,000 lost orders and roughly 1.6 million website errors, with Amazon’s Q AI tool among the contributors.
8. He says a March 5 Amazon outage caused a 99% drop in North American marketplace orders and 6.3 million lost orders.
9. Zitron argues that code written quickly without clear intent adds maintenance work, while citing Spotify’s CEO saying its top developers were “basically not writing code anymore.”
10. Software engineers, technology managers, and AI skeptics.
