liveBook · Manning

Alignment regression

As explained in Spec-Driven Development

The loss of previously achieved behavioral alignment as an implementation evolves and new features are added.

Alignment regression measures how often previously achieved alignment is lost as an implementation evolves and new features are added, harming behavior that used to work. section 3.4.2

Unlike time-to-aligned implementation, which measures how quickly a change first meets its intent, constraints, and verification evidence, alignment regression tracks whether alignment holds as the system grows. section 3.4.2

Figure 3.4 The SDD governance operating loop involves observing gaps between intent and implementation and correcting them by revising the specification and regenerating the implementation. We then attribute the gap in the specification to where the intent was lost in the SDD workflow to improve it, and we confirm the improvement is working when the same class of gap stops recurring.

The governance loop uses gaps between intent and implementation to guide specification revisions and regenerated implementations, then checks whether the same gap recurs. section 3.4.2

How Spec-Driven Development covers it

How deep it goes
Defines how often earlier alignment is lost and contrasts long-term retention with time-to-aligned implementation.
What you can do after
Evaluate whether earlier behavioral alignment persists as the implementation evolves and new features are added.
How it connects
Connects alignment regression to time-to-aligned implementation as a complementary measure of alignment durability.

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Spec-Driven Development. The section links point to the parts of the book it draws on.