随着大语言模型和 AI 代理在代码生成中的广泛应用，现有的代码协作平台正面临结构性挑战。[1]有评论指出，GitHub 及类似平台基于人工协作的工作流——分支、提交拉取请求、等待持续集成、代码审查、合并——已成为生产效率的主要瓶颈。[1]这是因为 LLM 和 AI 代理能够从简短的提示词生成完整功能，导致代码量和并行工作大幅增加，使得人类中心的审查流程无法适应机器级别的代码生成速度。[1]
现有基础设施的局限性在实际开发中表现明显。[1]例如，10 分钟的构建时间不再仅仅影响工程团队，而是对整个公司的交付效率产生阻滞。[1]为应对这一转变，业界应当从优化单个协作工具转向重构软件交付的底层基础设施。[1]这包括源控制、执行环境、工制品、缓存、身份认证和策略这六个核心原语。[1]观点认为，"未来十年的胜者不会构建更好的拉取请求，而是构建能让软件生成、验证和部署在机器规模上运作的基础设施"。[1]
GitHub's traditional collaboration model—based on branching, pull requests, code review, and merging—has become a bottleneck for contemporary software engineering, according to a perspective shared on Hacker News [1]. The shift is driven by the emergence of large language models and AI agents, which can generate complete functionality from minimal prompts and produce code at scales far exceeding what human-centric workflows can efficiently accommodate [1].
The current infrastructure exhibits critical constraints that extend beyond individual teams. Build times of ten minutes, for instance, now impact entire organizations rather than isolated engineering groups [1]. Under the human-centered paradigm, GitHub's sequential workflow—writing code in a branch, opening a pull request, waiting for continuous integration, undergoing review, and merging—struggles to handle the throughput demands of machine-generated code [1].
To address these challenges, the argument proposes a shift toward infrastructure designed for machine-scale operations [1]. Rather than optimizing the pull request experience, the suggested approach emphasizes six core infrastructure primitives: source control, execution environments, artifacts, caching, identity, and policy [1]. The contention is that competitive advantage in the coming decade will accrue to organizations that build infrastructure enabling software generation, validation, and deployment to operate at machine scale, rather than those merely refining existing human-oriented collaboration tools [1].