The semantic chaos of AI coding (and a proposed classification)

Author here. The AI coding space is semantically chaotic: "vibe coding," "AI-assisted," "context engineering," "augmented coding," "spec-driven", etc. Everyone's using different terms for overlapping concepts. Sometimes people

This article introduces what I call "the vibe coding trap" and attempts to bring clarity by proposing a standard classification: 4 distinct phases of AI adoption in software development.

- Phase 1: No AI (baseline)

- Phase 2: Autocomplete (Copilot-style)

- Phase 3: Conversational/Vibe Coding (for MVP/prototypes, otherwise painful)

- Phase 4: Structured AI (context engineering, specs-driven)

The terminology mapping alone might be helpful for teams trying to figure out "where are we?" and "where should we go?"

Curious: Are there any other initiatives to bring clarity about these terms?