Spec-Driven AI Development by Nate Hearns

​A Practical Workflow for Building with LLM Agents

​ Most AI coding starts with a prompt and a prayer. Spec-driven development offers a structured alternative: define what you want before asking an agent to build it.

​ This talk walks through the spec-driven pipeline and extends it with practices for building full-stack products with AI agents:

​ - Vision & Principles — project goals, constraints, coding standards, and how agents should work

​ - Specs — technical and product specifications of what you're building

​ - Use Cases — who you're building for and what gaps exist across features

​ - Tasks — units of work for agents, linked to specs

​ - Decision Records — when specs change, record why. Agents that don't know the reasoning will revert it.

​ - Tests & Benchmarks — the source of truth. Specs describe intent, tests prove behavior.

​ - Traceability — connecting specs to tasks to code so nothing drifts

​ We'll spec and build a small application live to show the workflow in action. Bring your laptop and an idea — we'll have time to try the process on your own project.

​ Who this is for: Developers who want to build real products with AI and want a repeatable process for doing it — from idea to shipped code.