The Standup with ThePrimeagen Whats really going on with AI, Expert weighs in
29 snips
Mar 13, 2026 Dimitri Spanos, an AI researcher with 20+ years focused on reliable, reproducible AI, and podcast cohost. He breaks down token costs and hardware realities. They debate whether developers will shift to reviewing AI-generated code. Conversation covers infrastructure limits, corporate incentives driving AI adoption, and practical meta-practices for safer integration.
AI Snips
Chapters
Transcript
Episode notes
Evaluate Bold Claims By Content And Timing Separately
- Separate claim content from timing when evaluating bold tech promises; the technical idea can be valid even if timelines slip.
- Dimitri used Musk's reusable rockets and Tesla FSD timelines as examples of delayed delivery.
Early AI Adoption Will Force A Review Heavy Workflow
- Expect a review-heavy phase where companies mandate AI-generated PRs and monitor token usage, making engineers primarily reviewers.
- Dimitri predicts businesses will link tokens to PRs and require seniors to sign off on junior AI outputs.
Avoid Rewarding Token Usage As A Raw KPI
- Beware organizational incentives that reward raw token usage; metric-driven mandates can force wasteful token spending.
- Dimitri noted managers may tie bonuses to token/PR increases, creating perverse incentives to inflate usage.