Deep Questions with Cal Newport Ep 386: Was 2025 a Great or Terrible Year for AI? (w/ Ed Zitron)
1227 snips
Jan 5, 2026 
In a thought-provoking discussion, Ed Zitron, an independent tech commentator and AI industry analyst, breaks down the chaos of AI in 2025. He delves into the controversy surrounding DeepSeek's low-cost training, critiques the overhyped idea of AI agents replacing jobs, and highlights the financial instability plaguing AI firms. They also question the impact of model scaling and the unrealistic expectations set for tools like GPT-5. Ultimately, Ed argues 2025 was a terrible year for AI, filled with hype and misleading narratives.
AI Snips
Chapters
Books
Transcript
Episode notes
GPT-5 Underwhelmed Despite Big Hype
- GPT-5 released to fanfare but underwhelmed many, revealing engineering trade-offs like router models that increased run-time cost.
- Ed Zitron's reporting uncovered router inefficiencies that most outlets missed.
GPT-5 Sparked Financial Scrutiny
- GPT-5's underperformance opened scrutiny into AI firms' finances and margins, spawning 'bubble' coverage.
- Public reporting began questioning whether companies could sustainably pay massive inference bills.
Anthropic's Massive Cloud Bills Revealed
- Ed obtained Anthropic's AWS bills showing multi-billion-dollar cloud spend over recent quarters.
- The disclosed spending contradicted narratives that some firms were significantly more efficient.
