← All Episodes
Based on Lenny's Podcast data
Lenny's Knowledge Sketch

This Is the
World Now

Boris Cherny
Engineer; TypeScript book author; Figma alum
FEB 19 2026
The Premise

AI-Augmented
Engineering Is Default

HUMAN CODEAI CODESHIPPED PRODUCT
"This is the world now. The question isn't whether to use AI to code — it's how to use it well."
  • AI writes 40-80% of code in productive teams today
  • Type systems + AI = fewer hallucinations, faster iteration
  • The review skill becomes as important as the write skill
  • Engineers who can't eval AI output are already falling behind
Framework

The AI-Augmented Dev Workflow

SPECAI DRAFTHUMAN REVIEW
faster first drafts
80%
acceptance rate with good specs
60%
reduction in boilerplate time
  • TypeScript + LLMs: type safety reduces AI hallucination rate significantly
  • Spec quality determines output quality — garbage in, garbage out
  • The 10-minute review beats the 3-hour write for standard patterns
  • Agentic workflows: let AI iterate, then review the diff
Boris's ruleWrite the test first. AI writes to your test. You review the diff. Ship.
Engineering Reality

Where AI Helps Most vs Least

Boilerplate95%API integration85%Test generation80%Complex logic45%Architecture25%Novel algorithms15%
The trap

Accepting AI output that works but is unmaintainable. Fast now, expensive forever.

The opportunity

AI handles the boring 70%. Engineers focus on the interesting 30% that requires real judgment.

Playbook

Engineer Better With AI

  • Invest in writing better specs — it's now your highest-leverage skill
  • Use TypeScript strictly — type safety is your AI guardrail
  • Review AI output with fresh eyes: does it match the intent, not just the prompt?
  • Keep a "prompt library" of patterns that consistently produce good output
The uncomfortable truthThe engineers who say "I don't use AI tools" are not principled — they're just slower.
Contrarian

Engineering Myths in the AI Era

AI can't write good codeINSTEAD →AI can't write good code without a good spec. That's a you problem, not an AI problem.
Code review is less importantINSTEAD →Code review is MORE important. AI is prolific but not always right.
Junior devs are obsoleteINSTEAD →Juniors who use AI are senior-level productive. Seniors who don't are junior-level competitive.
Ship fast, refactor laterINSTEAD →Ship thoughtfully. AI-generated code accumulates tech debt 3× faster if not reviewed.
𝕏︎ X / Twitterin LinkedIn📸 Instagram🔗 Copy link
0:00