Is “Vibe Engineering” Real Engineering or Just Terminology Inflation?

Simon Willison’s definition of “vibe engineering”

Willison introduced the term to separate the early, exploratory “vibe coding” style—where developers loosely interact with large language models—from a more experienced, responsible AI collaboration practice that can be deployed in real systems.

Three defining traits

Strong reliance on LLM generation and reasoning capabilities.

A highly interactive, improvisational, and informal development process.

The operator possesses enough experience to correct and back‑stop the system at critical points.

Although this approach is more mature than pure “vibe coding,” its maturity stems from individual expertise rather than systematic engineering; a skilled person can produce working software without formal methods.

Is it a new engineering paradigm?

The article notes that software history contains similar practices, but they are rarely labeled as a distinct engineering paradigm.

Prompt engineering and context engineering: truly engineering?

Prompt engineering focuses on probing model behavior with language, lacking a stable design space, reproducibility, and reusability.

Context engineering is mainly an experiential summary of what material to feed the model, rather than a rigorously structured model.

Both share common characteristics:

Heavy dependence on tacit knowledge.

Results tightly tied to the specific model version and time.

Absence of clear failure criteria and quality metrics.

These traits create natural tension with traditional engineering’s emphasis on predictability, verifiability, and inheritability, thereby diluting the meaning of “engineering” when applied indiscriminately.

What engineering truly requires

Explicit goal and constraint modeling.

Reusable methods and structures.

Verifiable, assessable, and evolvable outcomes.

By this standard, “vibe engineering” resembles a “pre‑engineering” state: it produces results that are hard to replicate, efficiency derives from personal skill amplification, and experience is difficult to institutionalize.

The real safety net

The viability of “vibe engineering” depends on existing software architecture, mature cloud infrastructure, established runtime and security boundaries, and implicit engineering constraints. Without these foundations, systems that rely solely on vibe, prompts, or context quickly become unmaintainable, unverifiable, and non‑evolvable at scale.

Conclusion

“Vibe engineering” is a thought‑provoking term that reminds us engineering can begin with intuition and exploration, but prematurely branding such practices as engineering blurs an important boundary. True engineering transforms vibe into structure and intuition into method, enabling repeatable creation under constraints.