The AI-Native Software Engineer
summary
Claims concrete: AI speeds scaffolding but stalls the 70% problem where the final mile needs senior work; Google data shows up to 21% more tasks completed per developer while PR review time jumps 91% and PR size up to 154%. Tools named include Cursor, GitHub Copilot, Gemini and Chrome DevTools MCP (agent sees the browser). Practices actually used: context engineering with memory banks and learnings files, spec-driven dev, and Trio programming as a workflow fix.
AI-written summary · about
chapters (14)
- 0:0001:34 | Intro: from "Will I be replaced?" to AI-Native Engineer
- 1:3404:23 | The shift: evolution from Implementer to "Orchestrator of agents"
- 4:2304:57 | The spectrum: "Vibe coding" vs. Managed engineering
- 4:5705:42 | A few words about JS Nation 2026 in Amsterdam
- 5:4207:39 | Mapping AI to the developer lifecycle (Inner vs. Outer Loop)
- 7:3908:09 | Google’s internal data: measuring real velocity gains
- 8:0910:44 | Why prompt engineering is dead (enter context engineering)
- 10:4414:27 | Practical pattern: Spec-driven development and "learnings" files
- 14:2715:39 | The trust gap: why 46% of devs distrust AI output
- 15:3917:15 | Tool reveal: Chrome DevTools MCP (AI seeing the browser)
- 17:1520:20 | The code review crisis: why PR times exploded by 91%
- 20:2021:44 | The "70% Problem": why the Last Mile is hardest for Seniors
- 21:4423:06 | Skill erosion and the threat to Junior hiring
- 23:0625:24 | Solution: "Trio programming" and No-AI challenges
related talks
26mBeyond the code: Leading engineering teams in the GenAI era
27mSoftware engineering with GenAI
15mThe AI-Augmented Engineer [...]
29mAI-Driven Development: Revolutionizing Productivity, Software Development Life Cycle
27mAI for software development: A reality check
26mAI Agents as Your Co-Developer
47mAI, Automation & DevEx: Fueling High-Velocity Engineering
14mAgent-Driven Software Development