Spec-Driven Development with AI Agents From High-Level Requirements to Working SW
summary
Spec-driven workflow uses explicit artifacts—requirements.md, plan.md, tasks.md—fed to AI agents so each step is reviewable and replayable. Demo uses JetBrains IDE integrations and OpenSpec to scaffold a CSV batch-import feature (MySQL, Testcontainers) and shows agents producing a plan of 156 lines vs 347 lines between models. Resulting code runs but has layering and transaction bugs until custom or verified skills and tighter specs are added and tests are reviewed.
AI-written summary · about
chapters (20)
- 0:00Introduction
- 1:20Spec-driven development
- 5:12Traditional SDD workflow
- 7:40Not waterfall development
- 9:06AI agents experiment
- 11:37Resilience with context workflows
- 13:23Markdown task tracking
- 14:30Transparent planning expectations
- 17:59Testing generated plan
- 20:07Progress tracking checkpoints
- 23:20Task granularity balance
- 26:30Pet projects to Spring apps
- 27:52Code quality review pitfalls
- 29:37Architecture improvement guidelines
- 31:47Custom vs off-the-shelf skills
- 32:09Toolkits for spec-driven workflows
- 34:40Live demo OpenSpec
- 43:35Future of spec-driven AI
- 45:01Success stories challenges
- 46:25Closing remarks Q&A
related talks
28mHow To Use Spec-Driven Development for Production Workflows
18mUsing Spec-Driven Development for Production Workflows
37mSpec Driven Development: AI Assisted Coding With Doc at the Core
30mSpec-Driven Evolution: How AI Agents Are Rewriting the OpenStack/K8s Pla...
45mSpec-Driven Development: the New Era of Development
01h 04mSpec-Driven Development: Agentic Coding at FAANG Scale and Quality
38mGet Shit Done with Get Shit Done - Spec-Driven development in AI times
25mDemystifying Spec Driven Development