LLMs are increasingly used to generate code, tests, and even architectural sketches. The problem is not their capability, but their lack of durable context: even when guided by ad-hoc prompts, their output is often inconsistent, unpredictable, and subtly misaligned with domain intent.
Specification-Driven Development (SDD) addresses this by treating specifications as more than just input for code generation. In SDD, specifications act as a shared, non-negotiable source of truth: a persistent memory that captures domain decisions, constraints, and language in a form that both humans and AI can consistently follow. Rather than asking a model to “do the right thing,” SDD makes domain intent explicit and enforceable within a bounded context.
Grounded in Domain-Driven Design, SDD reinforces a shared ubiquitous language, respects bounded contexts, and preserves the authority of domain experts by encoding their decisions into explicit, machine-readable rules. Through a series of hands-on exercises, participants will define domain specifications, evolve them collaboratively, and use them to guide AI agents in a controlled and deterministic way.
By the end of the session/workshop, participants will have produced a clear, domain-aligned specification and a concrete implementation plan derived from it.