SDAD Redefines SDLC for AI-Native Software Development

The advent of frontier coding agents with massive context windows is fundamentally reshaping the Software Development Life Cycle (SDLC). These agents can now ingest extensive functional requirement documents and repository contexts in a single workflow, making the quality of specifications paramount for autonomous delivery. This report introduces Spec-Driven Agentic Development (SDAD), a new paradigm that synthesizes disciplined upfront formalization with high-velocity agentic implementation. SDAD involves intent capture, machine-readable specification, agentic synthesis, and independent multi-agent verification, all under human sign-off. It re-evaluates the historical pendulum between Waterfall and Agile, positioning AI-code as a fourth production paradigm. The paper compares Human-Agile with Agentic-SDAD across artifacts, cadence, accountability, and security. It also extends the model to team role metamorphosis (e.g., engineers, QA, product), quantitative governance metrics (like Ambiguity Tax and Spec Fidelity), and pragmatic adoption strategies including hybrid estimation and staged migration. The core argument is that agentic speed doesn't eliminate engineering discipline but rather relocates it to upstream specification precision, explicit gates, and auditable provenance.