Reading is the New Writing in the AI Era
summary
Claim: developers spend 90% of time auditing LLM output because generated code compiles but hides anti patterns, deprecated APIs, and security gaps. The talk demonstrates the autopsy problem where on‑call engineers debug systems they never architected, lists concrete misses—race conditions, memory leaks, inefficient O(n) choices—and prescribes three pillars: data structures and complexity, concurrency and memory management, and system design plus threat modeling. Reading and verifying replaces blind acceptance.
AI-written summary · about
chapters (10)
- 0:00Welcome and Thesis
- 0:21Code Is a Commodity
- 0:56The Black Box Trap
- 1:53Skim vs Read Verify
- 2:36The Autopsy Problem
- 3:22What LLMs Miss
- 4:01Basics That Matter Now
- 4:59AI Assisted Mastery Path
- 5:35From Typists to Editors
- 6:17Final Call to Action
