The Circuit EP 147: Talking All Things AI with Benedict Evans
58 snips
Dec 29, 2025 
Benedict Evans, an independent technology analyst known for insightful analysis on major tech shifts, joins the discussion to unravel the complexities of AI today. He draws parallels between current AI development and the late-90s internet era, sparking debates over whether we are in an AI bubble. Evans emphasizes the necessity of productization for widespread user adoption, highlights differing impacts across industries, and compares Nvidia to Sun Microsystems in light of future challenges. Misunderstandings about AI's capabilities and compute demand also take center stage.
AI Snips
Chapters
Transcript
Episode notes
AI Is Big But Still Early
- AI today is like the internet in 1997 or smartphones in 2008: clearly huge but with many unknowns ahead.
- We can learn from past platform shifts but must accept significant uncertainty about winners and product shapes.
Productize AI, Don't Rely On Prompts
- Entrepreneurs must productize LLM capabilities into clear workflows rather than expecting users to craft prompts.
- Build UX, suggested actions, and domain thinking into interfaces so non-experts can use AI daily.
LLMs Need Workflow Guidance
- A GUI's role is more than removing keyboard commands; it guides users through workflow choices.
- LLMs must either dynamically generate UIs or be pre-wrapped with workflow knowledge to be useful.