Prototyping & the prototype-to-code boundary

AI makes it trivially easy to produce something that looks production-ready in minutes. That's the problem this dimension governs: not whether prototyping happens, but whether an organization has an explicit, consistent answer for what happens to a prototype the moment it looks good enough to ship.

Where most organizations start (Nascent)

Prototyping is ad hoc, and the boundary between prototype and production code is undefined or unenforced — meaning a prototype can become production code by accident, simply because nobody drew a line. The first real step is making that boundary explicit: label prototypes non-production by default, and establish a shared expectation that generated or exploratory code doesn't enter production just because it appears to work.

Where the real gains happen (Modeled → Integral)

The meaningful shift is creating a formal, documented decision point — evaluated against real criteria (architectural fit, test coverage, security implications) before any prototype code carries forward — and then moving that evaluation earlier, into construction itself. Once prototypes are built against the same architectural constraints the production path already uses, the frequency of "we built it, now we have to throw it away" drops, because the constraints that would have forced a rebuild were visible from the start.

What the top of the curve actually looks like (Telemetric)

At full maturity, specification, prototype, and test are produced as a single governed motion rather than separate phases. Because the prototype is part of that fused motion instead of a separate upfront investment, the carry-forward-versus-discard decision becomes immediate and genuinely low-cost either way: a prototype built purely to learn something can be discarded without losing the specification or test work behind it, and a viable one carries forward with both already in hand.

Why this dimension matters

This is the dimension where AI's biggest productivity gift — instant, plausible-looking output — becomes a real governance risk if left unmanaged. The capability being measured isn't "can you prototype fast," it's "do you have a real answer for what happens next," every time.

---

Drafted from the SDLC model's real locked content — including the transition and verification notes now folded into ai_native_sdlc_maturity_model.md itself (2026-07-27). D5 carries no open review flag.