Linear B adopts spec-driven development to accelerate AI engineering workflows
Linear B has transitioned its engineering team from ad hoc coding to a mandatory spec-driven workflow to manage AI-generated code. The methodology reduced the timeline for a complex feature from six months to less than one month.
Linear B's engineering team of approximately 30 engineers has transitioned from ad hoc 'vibe coding' to a mandatory spec-driven development workflow. The initiative began with a single interactive skill that generates specification documents for any story exceeding three story points, excluding DevOps tasks. This requirement is enforced through KPIs and weekly feedback loops. The company maintains two types of specs: frozen documents for completed features stored in Confluence, and live specs that continuously update to describe services.
Ariel Uluz and Ofer Afias argue that spec-driven development is essential for managing the volume of AI-generated code, shifting the engineer's role from writing code to reviewing and governing the output. The methodology reduced the timeline for a complex custom metrics builder feature from an estimated six months to less than one month, with two weeks dedicated to design and spec creation involving approximately 20 spec documents. The team also developed a marketplace of internal skills, including a troubleshoot skill that resolves 60-70% of customer support tickets before they reach engineering, and an implement skill currently in motion.