Vibe Coding While Live Coding - Lessons to Make AI Coding Safer
summary
Live coding example shows a concrete workflow: use three guardrails — source control with frequent commits, CI (GitHub Actions) gates, and automated tests — and run Playwright end-to-end checks like Playwright: 81 tests, 12 workers to catch regressions in ~20–30s, plus fast unit checks such as 430 unit tests in 12s. Agents (Squad/Gasteau) run parallel roles (Ripley, Hicks, Vasquez, Hudson) to build features, tests, and deploy streamermap.live.
AI-written summary · about
chapters (17)
- 0:00Vibe coding introduction
- 3:13What is vibe coding
- 4:19Live coding risks
- 7:50Need for guardrails
- 8:31Essential guardrails
- 11:19Source control safety
- 16:13Continuous integration as your calm adult
- 17:59Playwright insurance
- 21:29Unit tests feedback
- 22:53Agents for vibe coding
- 23:59Agent orchestration
- 28:14Vibe boldly build safe
- 30:01Q&A
- 31:05AI code failures
- 33:10Ban one habit
- 34:19AI fixes Playwright tests
- 35:21AI in CI/CD future
related talks
50mPanel: Beyond The Vibe: Code Quality First
33mEnsuring quality in an AI-driven world
38mWriting code live with AI
47mPanel about vibe coding
51mLive Panel Discussion: Less Code? More Vibes? Web Dev in the AI Era
01h 59mEmerging Patterns for Coding with Generative AI
01h 01m10 tips to level up your ai-assisted coding
30mBeyond the Hype: Using AI Coding Agents Today