⌘K
Change language Switch ThemeSign In
Narrow Mode
来自 GitHub 的数据表明，AI 工具正在创造“便利循环”，重塑开发者的语言选择 ==========================================
I InfoQ 中文 @InfoQ 中文
One Sentence Summary
The GitHub 2025 report reveals that AI coding tools are reshaping the language ecosystem through a 'convenience loop,' making strongly typed languages like TypeScript preferred over Python due to higher efficiency in AI collaboration.
Summary
Based on GitHub's 2025 Octoverse report, this article delves into how AI coding assistants are fundamentally changing developers' language choices. The core argument is that AI creates a 'convenience loop': AI enhances the ease of use for specific technologies (e.g., TypeScript), attracting more developers, which in turn generates more training data, making AI perform even better with that technology. Technically, strongly typed languages provide clear 'guardrails' for AI, effectively intercepting most compilation errors caused by LLMs. The report indicates that AI assistance has become a new constraint in technology selection, and new languages lacking large codebases face a vicious cycle of difficulty in gaining AI support. GitHub advises companies to shift their focus from the number of AI tool users to the actual output quality driven by AI.
Main Points
* 1. 'Convenience Loop' Drives Tech Ecosystem Evolution, AI Assistance Becomes a New Constraint in Developer Selection.AI enhances the ease of use for specific technologies, attracting more developers, which in turn generates more training data, making AI perform even better with that technology, creating a positive feedback loop. * 2. Strongly Typed Languages Are the Best Partners for AI Collaboration, Significantly Reducing Error Rates in AI-Generated Code.Static typed languages like TypeScript provide clear logical guardrails for AI. Research shows that 94% of compilation errors caused by LLMs are type check failures, and strongly typed systems can catch these errors before production. * 3. New Languages Face a 'Data Desert' Challenge in the AI Era, Struggling to Break Existing Ecosystem Monopolies.The essence of AI models is inference based on existing codebases. New languages lacking millions of lines of examples cannot receive effective Copilot support, leading to developer attrition and making it harder to accumulate training data. * 4. Enterprises Should Establish AI Output Metrics, Optimizing Team Prompts and Review Processes Through Data Dashboards.GitHub advises managers not to just count the number of AI tool users, but to focus on the quality of code produced by the tools, identifying the correlation between specific languages or models and defective code for targeted optimization.
Metadata
AI Score
84
Website mp.weixin.qq.com
Published At Today
Length 2026 words (about 9 min)
Sign in to use highlight and note-taking features for a better reading experience. Sign in now
作者 | Steef-Jan Wiggers
译者 | 平川
近日，GitHub 2025 年的 Octoverse 报告揭示了开发者可能没有意识到的一些事情。AI 编程助手不仅改变了开发者编写代码的速度，还影响了开发者最初选择的语言。
TypeScript 以 66% 的惊人年增长率成为 GitHub 上使用最多的语言，其原因不仅仅是框架的默认设置。GitHub 开发大使 Andrea Griffiths 称之为“便利循环（convenience loop）”，其工作原理是：当 AI 使某项技术变得方便使用时，开发者就会蜂拥而至，而这反过来产生了更多的训练数据，使 AI 在该技术上变得更加出色。
根据 GitHub 2025 年的 Octoverse 报告，到 2025 年 8 月，TypeScript 超越了 Python 和 JavaScript，成为 GitHub 上月活跃贡献者最多的语言，拥有 263.6 万开发者。这是十多年来最大的语言排名变动。当然，像 Next.js 和 Astro 这样默认使用 TypeScript 的框架提供了一些帮助。但对于为什么 TypeScript 与 AI 如此契合，有一个更深层次的技术原因。
（图片来源：GitHub 博客）
在最近的一篇博文中，Griffiths 进行了分析说明：
> 当一个任务或流程进行得顺利时，你的大脑就会记住。便利性吸引注意力。减少阻碍变成了偏好，而大规模的偏好可以改变生态系统。在 GitHub 上 ，80% 的新开发者在第一周内使用了 Copilot 。这些早期接触重新定义了“简单”的基线。
其技术优势显而易见。强类型语言为 AI 提供了明确的护栏。当你在 TypeScript 中声明 x: string 时，AI 立即就知道要忽略所有不适用于字符串的操作。JavaScript 采用的那种无拘无束的方法，对 AI 来说像是迷宫般的挑战。实际上，有研究支持这一点。Visual Studio Magazine 曾引用 2025 年的一项学术研究，由 LLM 造成的编译错误 94% 是类型检查失败。静态类型语言能在 AI 所犯的错误成为生产问题之前捕捉到它们。
TypeScript 并不是这个趋势中的孤例。GitHub 对类型化语言的分析显示，Luau（Roblox 的逐步类型化语言）年增长率为 194%，Typst（一个强类型的 LaTeX 替代品）增长了 108%。与此同时，当前有超过 110 万个公共存储库使用 LLM SDK。这已经成为主流，不再是实验性的了，并且集中在与 AI 协作良好的技术栈上。
Idan Gazit 领导着 GitHub Next 团队（ Copilot 背后的团队）。在另一次访谈中，他解释了 AI 如何从根本上改变了开发者在选择技术时的考量：
> 在 AI 技术出现之前，选择一种语言是在运行时、库生态系统和个人熟练度之间进行权衡。有了 AI 之后，出现了一个新的约束条件：如果我选择这种语言，模型会给我带来多少提升？
Python 仍然主导着 AI 项目开发；2025 年，GitHub 上新增的 AI 存储库有近一半开始时使用了 Python，这是因为它适用于模型训练和原型设计，而不是因为它是 AI 辅助应用开发的最佳选择。然而，若从整体发展态势来看，JavaScript/TypeScript 生态系统的发展规模远超其他任何领域。
Medium 博客作者 Cenk Çetin 分析 了这对整个行业来说意味着什么：
> 随着 AI 辅助编码的普及，提供静态类型检查的语言其地位不断上升。TypeScript 提供的严格类型系统有助于在 AI 生成的代码进入生产环境之前捕捉错误，增加代码可靠性。
Griffiths 希望团队能更有意识地考虑这个问题。她在 博文 中提出了一个简单的练习：
> 看看你最近做的三个技术决策：为新项目选择的语言，为新功能选择的框架，为工作流选择的工具。AI 工具支持在这些选择中占了多少比重？如果答案是“不多”，我敢打赌它比你意识到的更重要。
对于语言设计者来说，便利循环开创了一个具有挑战性的现实。在 GitHub 采访中，TypeScript 首席架构师 Anders Hejlsberg 直截了当地解释了这一点：
> AI 使用一种语言编写代码的能力与其见过的该语言的代码量成正比。它是一个大型复读机，辅以一定的推理能力。AI 已经看过大量的 JavaScript、Python 和 TypeScript 代码，所以它在编写这些语言的代码方面非常出色。从这个角度来说，新语言无疑处于劣势。
新语言陷入了恶性循环。Hejlsberg 指出，AI 模型基本上是复述它们之前见过的内容，并加入了一些推理结果。因此，如果你的语言没有数百万的代码示例，Copilot 就无法提供太大的帮助。而当无法获得 Copilot 的帮助时，开发者就会选择其他的东西。这也意味着你的语言永远不会有数百万行的代码示例。这是一个残酷的反馈循环，赢家已经锁定。
GitHub 的增长规模极为惊人：2025 年的 Octoverse 数据显示，平台拥有 1.8 亿开发者、6.3 亿个代码存储库，当年提交次数接近 10 亿次，同比增长率达 25%，相当于全年每秒新增一位用户。
对于试图理清这一切的领导者，Griffiths 给出了实用的建议：不要只统计使用 AI 工具的人数，要关注这些工具实际产出了什么。GitHub 新推出的 Copilot 使用指标仪表盘（企业版仍处于公开预览阶段）详细展示了谁在使用什么、使用的语言以及编码助手的采用方式。这有什么实际的意义吗？这有助于发现特定的语言或模型何时开始与存在缺陷的代码产生关联，从中可以看出团队需要优化提示词或加强代码审查的环节。
根据 Griffiths 的分析，可以得出的主要结论是：AI 兼容性正在悄然重塑你做出的每一个技术决策。在选择框架或语言时，你可能没有有意识地将其纳入考虑因素，但它就在那里。与 AI 助手不兼容的工具已经在失去市场。便利循环不在乎你的偏好，它只会使那些让编程感觉更轻松的东西加速发展。
原文链接： https://www.infoq.com/news/2026/03/ai-reshapes-language-choice/
I InfoQ 中文 @InfoQ 中文
One Sentence Summary
The GitHub 2025 report reveals that AI coding tools are reshaping the language ecosystem through a 'convenience loop,' making strongly typed languages like TypeScript preferred over Python due to higher efficiency in AI collaboration.
Summary
Based on GitHub's 2025 Octoverse report, this article delves into how AI coding assistants are fundamentally changing developers' language choices. The core argument is that AI creates a 'convenience loop': AI enhances the ease of use for specific technologies (e.g., TypeScript), attracting more developers, which in turn generates more training data, making AI perform even better with that technology. Technically, strongly typed languages provide clear 'guardrails' for AI, effectively intercepting most compilation errors caused by LLMs. The report indicates that AI assistance has become a new constraint in technology selection, and new languages lacking large codebases face a vicious cycle of difficulty in gaining AI support. GitHub advises companies to shift their focus from the number of AI tool users to the actual output quality driven by AI.
Main Points
* 1. 'Convenience Loop' Drives Tech Ecosystem Evolution, AI Assistance Becomes a New Constraint in Developer Selection.
AI enhances the ease of use for specific technologies, attracting more developers, which in turn generates more training data, making AI perform even better with that technology, creating a positive feedback loop.
* 2. Strongly Typed Languages Are the Best Partners for AI Collaboration, Significantly Reducing Error Rates in AI-Generated Code.
Static typed languages like TypeScript provide clear logical guardrails for AI. Research shows that 94% of compilation errors caused by LLMs are type check failures, and strongly typed systems can catch these errors before production.
* 3. New Languages Face a 'Data Desert' Challenge in the AI Era, Struggling to Break Existing Ecosystem Monopolies.
The essence of AI models is inference based on existing codebases. New languages lacking millions of lines of examples cannot receive effective Copilot support, leading to developer attrition and making it harder to accumulate training data.
* 4. Enterprises Should Establish AI Output Metrics, Optimizing Team Prompts and Review Processes Through Data Dashboards.
GitHub advises managers not to just count the number of AI tool users, but to focus on the quality of code produced by the tools, identifying the correlation between specific languages or models and defective code for targeted optimization.
Key Quotes
* Convenience attracts attention. Reducing friction becomes preference, and large-scale preference can change ecosystems. * Before AI technology, choosing a language was a trade-off between runtime, library ecosystem, and personal proficiency. With AI, a new constraint has emerged: if I choose this language, how much will the model enhance my capabilities? * AI's ability to write code in a language is directly proportional to the amount of code it has seen in that language. It is a large parrot, supplemented by some reasoning capabilities. * The convenience loop doesn't care about your preferences; it only accelerates what makes programming feel easier.
AI Score
84
Website mp.weixin.qq.com
Published At Today
Length 2026 words (about 9 min)
Tags
GitHub Octoverse
TypeScript
AI Coding Assistants
Programming Language Trends
Static Typing
Related Articles
* OpenAI Frontline Development Observations: Those Who Can Manage 10-20 Agents Simultaneously and Run Hour-Long Tasks Are Leaving Other Engineers Far Behind * From Context to Long-Term Memory: Architectural Design and Practice of LLM Memory Engineering\" architecture.") * 【User Behavior Monitoring】Beyond Just Using Tools: A Step-by-step Guide to Building a Frontend Event Tracking SDK * “AI on the Front Lines: How Developers are Reshaping the Software Development Process” | Roundtable Discussion * 1,500 PRs, 0 Human Coders: Building a Million-Line Internal Product Driven by Codex * [[Performance Monitoring] Stop Being Just a Tool User! A Step-by-Step Guide to Writing a Frontend Performance Monitoring SDK](https://www.bestblogs.dev/en/article/3e3a598c "This article provides a detailed guide on building a frontend performance monitoring SDK from scratch, covering core metric capture, layered architecture design, data processing/reporting, and engineering-driven publishing to empower developers to escape the \"package-caller\" dilemma.") * #450. The 'Printing Press' Moment for Programmers: A Conversation with Claude Code Creator Boris Cherny on the Evolution of AI Programming * 【Issue 3625】Writing TypeScript Doesn't Guarantee Safety: Boundary Design is Key * Practices and Reflections on Vibe Coding in Code Generation and Collaboration * After selling his previous company, this serial entrepreneur secured $16 million to disrupt AI agent development with open source HomeArticlesPodcastsVideosTweets
Data from GitHub Shows AI Tools Are Creating a 'Convenien... ===============