Spec-Driven Development
Spec-Driven Development is an AI programming methodology in which detailed specification documents are written before an AI agent begins execution. This approach decouples "defining what" from "implementing how," constraining and guiding the agent's behavior to improve intent fidelity and reduce the disorder associated with unsupervised AI-generated code.
Timeline (last 90 days)
Setting up a spec-driven development process takes about half a day, the guidance is generic, with part of the time spent adapting it to software development and part spent formulating your own rules.
In Spec-Driven Development, the agent's use of Conventional Commits means that every commit carries semantic information, making it easier for humans to review and trace rule changes.
GitHub Spec Kit describes Spec-Driven Development as an 'inversion': in traditional development specifications serve the code, whereas in Spec-Driven Development the code serves the specifications.
Spec-Driven Development is a development methodology that treats software specifications as the single source of truth, emphasizing writing the specification first and then generating code, with the code not to be edited manually.
The pauses between phases in spec-driven development are an intentional safety gate/checkpoint design
The essence of SDD is decoupling the specification that describes 'what to do and why' from the implementation of 'how to do it'.
SDD compensates for the stateless nature of models by using persistent specification documents as external memory.
The three major benefits of SDD are controlling large changes with small modifications, eliminating context decay, and improving intent fidelity.
In SDD, each feature follows a repeatable cycle of plan → implement → validate and is developed on an independent branch.
SDD is also applicable to existing brownfield projects, where agents can reverse-engineer SDD artifacts by exploring the existing codebase.
9 more timeline events
All Facts (20)
GSD的规范驱动开发遵循三阶段工作流：规范起草、规范审查和规范执行
90%UnverifiedSDD（Spec-Driven Development，规范驱动开发）在日常小迭代中写Spec的成本远高于直接写代码，且Spec基本是一次性的难以积累价值
70%UnverifiedSetting up a spec-driven development process takes about half a day, the guidance is generic, with part of the time spent adapting it to software development and part spent formulating your own rules.
50%UnverifiedSpec-Driven Development is a development methodology that treats software specifications as the single source of truth, emphasizing writing the specification first and then generating code, with the code not to be edited manually.
50%UnverifiedGitHub Spec Kit describes Spec-Driven Development as an 'inversion': in traditional development specifications serve the code, whereas in Spec-Driven Development the code serves the specifications.
50%UnverifiedIn Spec-Driven Development, the agent's use of Conventional Commits means that every commit carries semantic information, making it easier for humans to review and trace rule changes.
50%UnverifiedThe pauses between phases in spec-driven development are an intentional safety gate/checkpoint design
50%UnverifiedSDD is also applicable to existing brownfield projects, where agents can reverse-engineer SDD artifacts by exploring the existing codebase.
50%UnverifiedThe essence of SDD is decoupling the specification that describes 'what to do and why' from the implementation of 'how to do it'.
50%UnverifiedSDD compensates for the stateless nature of models by using persistent specification documents as external memory.
50%UnverifiedThe three major benefits of SDD are controlling large changes with small modifications, eliminating context decay, and improving intent fidelity.
50%UnverifiedIn SDD, each feature follows a repeatable cycle of plan → implement → validate and is developed on an independent branch.
50%UnverifiedThe Constitution consists of three pillars: Mission, Tech Stack, and Roadmap
50%UnverifiedSDD advises developers to review small changes with a human-in-the-loop, and to let agents modify files rather than editing them manually, in order to keep all artifacts in sync and avoid drift
50%UnverifiedThe basic idea of SDD (Spec-Driven Development) is to fully describe requirements, architecture, interface contracts, and test expectations in structured documents before AI writes code, and then have AI generate code step by step according to the specification.
50%UnverifiedThe spec-first philosophy aligns with the rising industry trend of Spec-Driven Development.
50%UnverifiedThe feature development loop follows three steps—plan, implement, and verify—with each feature isolated on an independent branch.
50%UnverifiedThe SDD workflow is applicable to both Greenfield (brand-new projects) and Brownfield (existing codebases) projects.
50%UnverifiedThe three core advantages of SDD are: controlling large-scale code changes with small modifications, eliminating context decay between sessions, and improving intent fidelity.
50%UnverifiedThe SDD workflow is divided into two levels: the project layer (establishing a "constitution" that defines immutable standards) and the feature layer (the feature development loop).
50%