The New SDLC With Vibe Coding
26 Jun 2026
It is very well written, with almost no fluff.
If you want to get a very high-level foundation for understanding how AI is currently being used in engineering, it is worth reading.
If you want to get a very high-level foundation for understanding how AI is currently being used in engineering, it is worth reading.
Key insights:
- The engineer becomes an arbiter of quality. They help the agent by defining what system correctness means and setting the boundaries for verification.
- Vibe coding and agentic engineering are two parts of the same spectrum. Modern software development is drifting from one extreme to the other.
- The quality of generated code depends on the context provided, not on how clever the prompts are.
- 80% rule. Vibe coding is good for prototypes (getting the first 80%), but it scales poorly afterward (doing the rest of 20% of work) and becomes very expensive. Moreover, vibe coding along is a pathological generator of massive technical debt.
- Tests are becoming one of the ways to communicate with an AI system.
- Traditional tests are not enough. We also need to verify the agent’s intermediate reasoning while it is working.
- AI-based development is not only about the model. It is the model plus the surrounding harness.
- The engineer becomes a conductor or orchestrator of AI agents.
- The key skill in the coming years will not be the ability to write more code, but the ability to manage agents effectively.
- When adopting AI, we should not think only about “generating code faster.” We should think about the total cost of ownership. Because tokens can be burned very easily.
Oleksandr Romanov
Senior Software Engineer in Test
He/him
14+ years in testing and engineering | Co-host at Testing Minutes podcast | Writing and podcasting about test engineering, performance, AI, blockchain, and distributed systems | Views are my own
Sign in
to comment
Better than a generic video, see YOUR test, live, ready to show you what matters most: quality at scale.
Explore MoT
Thu, 1 Oct
A tech conference to help you navigate the ever-shifting landscape of Quality Engineering, AI, Leadership, Product, Accessibility and Security.
Advanced prompting skills to turn AI into your trusted testing companion.
Into the MoTaverse is a podcast by Ministry of Testing, hosted by Rosie Sherry, exploring the people, insights, and systems shaping quality in modern software teams.
