Productivity: a dialogue
summary
Clear examination of whether generative AI actually speeds up work: cites a 2025 result where developers perceive themselves 20% faster but are 20% slower, shows Honeycomb’s adoption as an example of worker‑led tooling (Honeycomb generative AI adoption case), and digs into concrete harms like deskilling and the human‑in‑the‑loop accountability trap. Also covers economics (profitable inference vs costly training, Q2 2026 profitability caveats) and practical artifacts like side projects and verification burdens, so expect concrete claims, numbers and tradeoffs.
AI-written summary · about
chapters (16)
- 0:00Productivity dialogue intro
- 1:47Agile manifesto and participation
- 5:00Half-assed agile adaptations
- 6:45Generative AI productivity myths
- 8:10Deskilling and cognitive offloading costs
- 10:1610x devs and technical debt
- 12:21LLMs breaking code review
- 13:23AI productivity perception gaps
- 14:26AI service providers' economic fragility
- 16:53Tech bubble and job risks
- 17:58AI deskilling scenarios
- 19:43Accountability risks in AI
- 21:48AI exposure and societal harms
- 23:56Engineering discipline with non-deterministic systems
- 25:44Social license and AI environmental costs
- 26:49Industry tensions and dialogue need
related talks
31mReflections of AI: A Trilogy in 4 Parts
26mGenerative AI: Productivity or Threat for Developers
39mBuild without limits: lessons learned from using generative AI dev tool
42mUX+AI: Mastering the UX of AI-Powered Products
45mHyper Innovation in Exponential Times
38mFrom code to culture: How AI is reshaping the engineering organization
23mThe AI productivity paradox: Why feeling faster isn't the same as going faster
18mHumans + AI: The Future of Software Development Teams