Goodbye Claude, Hello Pi
821 points
• 4 days ago
• Article
Link
该作者在使用 Claude Code 超过一年后，决定不再将它作为主要的开发工具。尽管他承认该工具在推动当前 AI 辅助软件开发浪潮方面具有重要意义，但接连不断的技术挫折让他最终转向 Pi 。他表示，即便是近期的更新（例如对 AGENTS.md 文件的支持）也来得太晚，无法弥补他对产品方向日益积累的不满。
他认为，争议的一大焦点在于 VSCode 内的使用体验。作者提到一个限制：一次只能有一个会话连接到 IDE，这导致在具备正确的 IDE 文本选择支持的情况下，无法同时运行多个任务。为了解决这个问题，他尝试使用官方的 Claude Code GUI 扩展，但遇到了损坏的斜杠命令以及杂乱的消息界面。稳定性问题再加上他对产品团队把"快速、面向企业的功能堆砌"放在核心功能之上的看法，让整体体验越来越难以承受。
作者还担心 Claude Code 的更新频率过高，称这些更新经常打乱个人的配置。一年内发布了数百次改动，其中许多会调整系统提示词或工具描述，进而让 AI 的行为变得不可预测，并且频繁与既有项目规范相冲突。他认为当前的系统提示词包含了过多、甚至不必要的冗余内容，例如对模型本身已经能够独立完成的任务进行说明，从而浪费大量计算资源。
综合这些问题，作者认为 Claude Code 缺乏清晰且一致的产品愿景。他推测，如果一种"在缺少足够反对意见的情况下不断加功能"的文化占了上风，就可能让产品在体验上与高阶用户的真实需求产生脱节。他强调，尽管打造这类工具本身就很难，但由于缺少细致的产品取舍，新功能往往会拉低整体体验。
与之相对，作者认为转向 Pi 是一次积极的变化，主要体现在其稳定性、极简的设计以及更强的可定制性。通过切换到一个能让用户拥有更多个人控制权、并支持构建扩展的工作底座，他成功复现并改进了 Claude Code 中原本缺失的能力。最后，他鼓励同样被这些问题困扰的人尝试开源替代方案，并指出当前 AI 编程工具的生态仍在持续演进，优先满足自己的工作流程需求尤为关键。
The author has decided to stop using Claude Code as their primary development harness after more than a year of use. While they acknowledge the tool's significance in pioneering the current era of AI-assisted software development, a series of persistent technical frustrations led to the decision to migrate their workflow to Pi. The author notes that even recent updates, such as the support for AGENTS.md files, arrived too late to address their accumulated dissatisfaction with the product's direction.
A major point of contention involves the user experience within VSCode. The author describes a limitation where only one session can be connected to the IDE at a time, preventing multiple tasks from running concurrently with proper IDE text selection support. When attempting to use the official Claude Code GUI extension to solve this, the author encountered broken slash commands and a cluttered message interface. These stability issues, combined with the perception that the product team prioritizes rapid, enterprise-focused feature bloat over core functionality, became increasingly difficult to manage.
The author expresses concern over the high frequency of updates to Claude Code, which they claim often interfere with personal setups. With hundreds of releases occurring in a single year, many of which alter system prompts or tool descriptions, the AI's behavior becomes unpredictable and frequently contradicts established project conventions. They argue that the current system prompts contain excessive, unnecessary overhead, such as instructions for tasks that the models are already capable of performing independently, which the author believes wastes significant compute resources.
Reflecting on these challenges, the author suggests that Claude Code suffers from a lack of cohesive product vision. They suspect that a culture of adding features without sufficient pushback has resulted in a product that feels disconnected from the needs of individual power users. They emphasize that while building these tools is inherently difficult, the lack of careful product curation creates an environment where new additions often degrade the overall experience.
In contrast, the author finds the transition to Pi to be a positive shift, citing its stability, minimalist design, and superior customizability. By moving to a harness that allows for more personal control and extension building, the author has managed to replicate and improve upon the functionalities that were lacking in Claude Code. They conclude by encouraging others facing similar frustrations to explore open-source alternatives, suggesting that the current landscape of AI coding tools is still evolving and that prioritizing one's own workflow requirements is essential.
493 comments • Comments Link
. 许多开发者正从基于终端的（TUI）界面，转向用于与 LLM 交互的专用 GUI 桌面应用。原因在于后者在历史记录管理、富文本渲染、集成图像支持以及会话组织等方面更清晰、更顺畅。
. 尽管 TUI 倡导"高活动密度"的工作流效率，并强调通过 SSH 远程运行 agent 的能力，但不少人发现，与现有的现代 GUI 替代方案相比，基于终端的 harness 往往更容易出现滚动错误、源代码导航体验较差，而且多任务处理能力也有限。
. 用户自建的定制化 harness 已成为重要趋势。开发者可以据此打造量身定制的工作流，控制工具执行方式，手动管理上下文，从而避免被大型 SaaS 提供商"固化"的设计所限制。
. 像 Pi 这样的项目，是个人 coding agents 的热门且可扩展基础。它让开发者能够构建自定义的 Web 或移动端前端，并集成诸如定时任务、沙盒，以及特定领域的压缩逻辑等更具特色的功能。
. Anthropic 的官方 Claude Code 因集成能力强、模型表现好而广受认可，但用户经常抱怨其内部机制不够透明、系统提示词较为强势，以及对特定工作流的限制过于严格。
. 订阅价值仍然是选择 harness 的核心驱动因素：有人留在 Claude Code 或 Cursor，以尽可能多地访问 frontier 模型；也有人选择基于 API 的工作流，以保留模型与 harness 的灵活性。
. "harness 对 UI"的争论，本质上是在深度、持久上下文的需求（例如管理海量代码库）与希望获得无缝体验、从而尽量减少手动配置摩擦之间的拉扯。
. 越来越多的人倾向于采用有效的 agent 编排方式，例如把任务拆成子程序，或使用"移交"形式的 markdown 文件，而不是一味依赖自动压缩。许多人发现，自动压缩在长期项目中往往不可靠，甚至可能导致数据丢失。
. 安全与沙盒是关键考量。成熟用户正在绕开简单的"自动模式"分类器，转而使用诸如 Docker 或 Bubblewrap 之类的容器化环境，确保 agent 与生产代码的交互安全、可控。
. 尽管工具发展迅速，但仍有开发者提醒不要过度定制（"插件疲劳"）。他们建议，随着模型不断进步，重点应放在简单、稳健的 UX 上，而不是无休止地维护复杂的、自建 agent 基础设施。
. 开发社区目前在 AI 公司提供的高性能"frontier"类 harness，与高度个性化、由用户自建的 agent 架构之间存在分歧。尽管官方的 GUI 与 TUI 工具能带来即时可用性，但许多高阶用户正在选择构建定制化封装，以更好地掌控上下文管理、安全性以及面向特定项目的工作流。归根结底，这些工具之间的取舍，通常是在集成式 SaaS 功能带来的便利性与长期拥有个人编码 harness 的灵活性之间做权衡。 • Many developers are transitioning from terminal-based (TUI) interfaces to dedicated GUI desktop applications for LLM interaction, citing benefits such as cleaner history management, rich text rendering, integrated image support, and easier session organization.
• While TUI enthusiasts argue for the efficiency of "activity-dense" workflows and the ability to run agents remotely over SSH, others find terminal-based harnesses to be prone to scrolling bugs, poor source-code navigation, and limited multitasking capabilities compared to modern GUI alternatives.
• Custom, user-built harnesses have emerged as a significant trend, allowing developers to create tailored workflows, control tool execution, manage context manually, and avoid dependence on the "opinionated" design of large SaaS providers.
• Projects like Pi serve as popular, extensible foundations for personal coding agents, enabling developers to build custom web or mobile frontends and integrate unique features like scheduled tasks, sandboxing, and domain-specific compaction logic.
• Anthropic's official Claude Code is widely recognized for its powerful integrations and model performance, but users often express frustration with its opaque internals, forced system prompts, and rigid adherence to specific workflows.
• Subscription value remains a core driver for harness selection, as some users stay with Claude Code or Cursor to maximize access to frontier models, while others opt for API-based workflows to maintain model and harness flexibility.
• The "harness versus UI" debate is often a tension between the need for deep, persistent context (managing massive repositories) and the desire for a seamless user experience that avoids the friction of manual configuration.
• Effective agent orchestration, such as breaking down tasks into sub-routines or using "hand-off" markdown files, is increasingly favored over relying on automated compaction, which many find unreliable or prone to data loss in long-running projects.
• Security and sandboxing are critical concerns; sophisticated users are bypassing simple "auto-mode" classifiers in favor of containerized environments like Docker or Bubblewrap to ensure safe, controlled agent interactions with production code.
• Despite the rapid evolution of tools, some developers warn against excessive customization ("plugin fatigue"), suggesting that as models improve, the focus should remain on simple, robust UX rather than endlessly complex, self-maintained agent infrastructures.
The development community is currently divided between high-performance "frontier" harnesses provided by AI companies and highly personalized, user-built agent architectures. While official GUIs and TUI tools offer immediate usability, many power users are opting to build bespoke wrappers to gain better control over context management, security, and project-specific workflows. Ultimately, the choice between these tools often balances the trade-off between the convenience of integrated SaaS features and the long-term flexibility of owning one's coding harness.