Vibe Testing Is a Disaster Waiting to Happen

Vibe testing: the practice of letting AI write your code and then letting AI test it. Sounds efficient. Isn't. The data on AI-generated code quality is genuinely alarming - and most teams haven't seen it. In this video I go through what the research actually says, why "the AI tested it" is not a testing strategy, and what teams should be doing instead. 📊 Sources and data referenced: CodeRabbit — State of AI vs Human Code Generation Report (470 pull requests analyzed) State of AI-Generated Code 2026 (21.6M lines scanned, 346,944 issues found) SmartBear 2026 Survey (46% shipped AI code that failed in production, 69% still confident) New Relic 2026 State of AI Coding (bugs per developer up 54%, incidents up in 78% of orgs, senior engineer rework up in 86%) ICSE 2026 DeepTest research (932,000+ AI-authored pull requests, 80% with weak or no oracle signals) 🎧 Spotify: https://open.spotify.com/show/1Xdh0RP... 🤝 LinkedIn:   / ciepielewski   💻 Website: https://testeroni.org/ ✅ Subscribe:    / @testeroniorg   ⏱️ Timestamps: 00:00 — Intro 00:22 — What is vibe testing? 01:25 — The core problem: AI grading its own exam 03:01 — The numbers that should scare you 04:33 — 46% shipped broken AI code, 69% still confident 05:15 — "That's not confidence. That's a coping mechanism." 05:51 — What's actually happening under the hood 06:30 — What's an oracle (and why 80% of AI tests lack one) 07:06 — Green dashboards are lying to you 09:40 — What to actually do about it 12:04 — This is a management problem, not a QA problem 12:43 — No neat optimistic ending #VibeTesting #QA #SoftwareTesting #AITesting #TestAutomation #QualityAssurance #Testeroni