Vibe Was Never the Problem. But the Missing Half Starts Before the Build.
Why vibe coding needs software engineering before we let the agents loose. Don's loop gets the post-build part right. With coding agents, though, I think a lot of the damage can happen before we ever get there. In Vibe Was Never the Problem: The Missing Half of Vibe Coding , Don Johnson argues that intuition isn't what goes wrong in vibe coding. Stopping at intuition is. His fix is a loop: Vibe →…
Don Johnson's book "Vibe Was Never the Problem: The Missing Half of Vibe Coding" argues that the problem with vibe coding is not the coding itself, but rather the missing build and understanding phases. Johnson suggests a loop: Vibe → Build → Break → Understand → Stabilize → Perfect, emphasizing that intuition alone is insufficient in this process.
The author agrees with Johnson's middle portion of the loop, but proposes a different sequence. The author suggests adding a Spec step before the build: Vibe → Spec → Break the Design → Build → Break the Build → Understand → Stabilize → Perfect. This addition addresses the rapid pace at which coding agents can generate code, leading to potential architectural debt if not properly reviewed.
The author explains that a lightweight spec at the beginning of the process changes the division of labor. Instead of AI designing and building, followed by a human discovering what happened, the process becomes: Human defines intent and boundaries → AI builds → human verifies. This step helps prevent architectural debt and ensures that the system aligns with the intended use case.
The author also emphasizes the importance of breaking the design before implementation becomes expensive. By breaking the design, the team can attack assumptions, consider worst-case scenarios, and identify trust boundaries, all of which are crucial in the early stages of development. The second break, breaking the build, involves testing, fuzzing, injecting failures, and subjecting the system to various stresses to ensure it is implemented correctly.
While the author agrees that building before the spec can be beneficial for discovery, they argue that there must be a spec before the system can be shipped. Without a defined spec, the team risks creating architectural debt and relying on opinions rather than well-defined requirements. The spec, which is a one-page document, serves as a crucial checkpoint to ensure that the system aligns with the intended users, boundaries, requirements, and success criteria.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
