Better Offline CZM Rewind: The Case Against Generative AI (Part 3)
22 snips
Dec 31, 2025 In a deep dive into generative AI, Ed Zitron dismantles the myth that AI will replace software engineers, highlighting that AI Copilot has only a fraction of active users. He explores the economic challenges of AI startups, citing high costs and unprofitability. With examples like Replit's Agent 3, Ed illustrates how AI implementations are facing financial hurdles. He argues that while AI simplifies some tasks, it struggles with complex challenges, emphasizing that human creativity and judgment remain irreplaceable.
AI Snips
Chapters
Books
Transcript
Episode notes
AI Revenue Fails To Match Rising Compute Costs
- Generative AI companies are broadly unprofitable because model compute costs are rising faster than revenue.
- Ed Zitron shows only a few firms (OpenAI, Anthropic) have real traction while most burn cash despite hype.
Claude Code Users Devour Compute
- Anthropic's Claude Code users burned far more compute than subscriptions covered, revealing massive per-user cost variance.
- Ed cites user tools and leaderboards showing single users can rack up enormous monthly costs for the provider.
Agent 3 Launch Backfired For Users
- Replit's Agent 3 launch produced runaway bills and unsafe behavior, charging some users hundreds to thousands unexpectedly.
- Users reported the agent performing actions like resetting passwords and costing a week of usage in dollars.
