How I Fight AI Brain Fry
313 points
• 4 days ago
• Article
Link
使用像 Claude Code 或 Codex 这样的 AI 代码工具，可能会迅速引发认知负荷过载。当用户同时应付多个 terminal 标签页、后续任务以及不断切换的情境时，往往会走到精神耗尽的临界点：注意力被打散，原本的目标也会因为打开的线程太多而变得模糊。作者把这种现象称为 AI brain fry，它源于同时管理过多的序列与上下文加载的困难。
为了解决这个问题，作者把工作流管理当作一个优先队列（priority queue），而不是一连串缺乏结构的聊天（chat sessions）。通过把项目跟踪从临时的聊天窗口移到仓库本身，作者确保即便某次会话（session）中断，也不会丢失任何关键信息。每个任务都被定义为一个放在相关代码旁边的 markdown 文件，项目当前状态通过 git commits 记录下来。这会形成一个持久的外部记录，把"追进度"的负担从人的大脑里移除。
该系统依赖一种"代理式"（agentic）的工作流：任务按周期（cycles）推进。代理负责调研、非敏感构建和验证，但被严格禁止进行最终的生产部署或做出关键决策。代理遇到需要人类判断的卡点时，会把该事项放入一个集中式的 owner inbox 。随后，一个脚本会把整个看板（board）汇总成一个单一的 Claude artifact，为用户生成一页更易审阅的页面，清楚说明构建了什么以及原因。这样就不再需要为还原一次变更而在终端日志里反复滚动查找。
这套方法的关键在于"收尾"（closure）的仪式感。通过把每个任务隔离开来，用户只需为处理某一个事项加载必要的上下文，完成决策或发起需要的操作，然后立刻把该票据（ticket）退回队列以关闭它。关闭票据是最重要的一步，因为这意味着该事项不再占用"持续占用"的心理内存（mental RAM）。用户就能在清醒的状态下切换到下一个事项，从而保持一次只解决一个具体问题的专注。
尽管作者也承认这项工作仍然很费劲，但这种方式确实能避免被分心或迷失。当前最大的难点在于完善工具链，让上下文尽可能在一分钟内就能被快速消化。通过防止票据在等待人类输入时陷入无休止的循环，作者打造了一套"活的流程"（living process），在有效管理与 AI agent 交互带来的心理压力的同时，显著提升了生产力。
Working with AI coding tools like Claude Code or Codex can quickly lead to cognitive overload. When users juggle multiple terminal tabs, follow-up tasks, and shifting contexts, they often reach a point of mental exhaustion where focus shatters and the original goal becomes obscured by the sheer number of open threads. This phenomenon, which the author describes as AI brain fry, stems from the difficulty of managing too many sequences and context loads simultaneously.
To combat this, the author treats workflow management like a priority queue rather than an unstructured series of chat sessions. By moving project tracking out of transient chat windows and into the repository itself, the author ensures that no critical information is lost if a session terminates. Every task is defined as a markdown file located directly alongside the relevant code, with the current state of the project captured through git commits. This creates a persistent, external record that removes the burden of tracking progress from the human brain.
The system relies on an agentic workflow where tasks are processed in cycles. The agent handles research, non-sensitive builds, and verification, but is strictly prohibited from making final production deployments or critical decisions. When the agent encounters a hurdle that requires human judgment, it places that item into a centralized owner inbox. A script then compiles the entire board into a single Claude artifact, providing the user with a streamlined review page that explains exactly what was built and why. This eliminates the need to scroll through terminal logs to reconstruct a change.
The core of the strategy lies in the ritual of closure. By isolating every task, the user loads only the necessary context for one item, renders a decision or requests an action, and then immediately closes that ticket by returning it to the queue. Closing the ticket is the most important step, as it signals that the item no longer requires active mental RAM. This allows the user to transition to the next item with a clear head, maintaining focus on one specific problem at a time.
While the author admits that the work remains demanding, this method successfully prevents the feeling of being distracted or lost. The primary challenge currently lies in refining the tooling to ensure that context can be digested quickly, ideally within a minute. By preventing tickets from becoming stuck in endless loops while waiting for human input, the author has created a living process that significantly improves productivity while effectively managing the mental strain of interacting with AI agents.
215 comments • Comments Link
- Productivity 目前正处于一个转型期：被重新定义。 AI 模型既需要持续互动才像样，又慢到无法实现真正的单任务深度处理，于是很多人选择同时跑多个 agentic sessions 来避开"空闲"时间。
- 越来越多的人达成共识：以"agent-heavy"为主的工作流，往往会模仿 call center 那种有毒且高压的环境——在那里，成功取决于持续产出，而不是认知深度，或可持续的高质量产出。
- 同时管理多个 autonomous agents 会带来一种"上下文切换"的隐性成本：注意力被稀释，代码里的方案变得脆弱、拼贴感很强，最后反而增加了人类开发者的维护负担。
- 也有一些 developers 已经取得效果，采用"manager"式的调度：用 Kanban boards 和基于 git 的 issue tracking 把自己当成 supervisor，负责分派任务与审查逻辑，而不是亲自手动实现一切。
- "AI brain fry" 被视为一种真实的心理负担：表现为 deep-work 能力的衰退，以及不自觉地模仿 LLMs 那种断续又过于正式的写作腔调。很多人觉得这种风格不够人性，读起来也令人疲惫。
- 防止 burnout 的最有效解药，可能是刻意回归以人为本的习惯：手动 coding 、阅读 books 、拥抱空闲时间，而不是本能地再开一个新的并行任务。
- 用于 agent 管理的 sophisticated tooling 正变得越来越必要。许多 developers 在打造自定义界面，或使用与 git 集成的项目看板，以获得做出明智决策所需的 context，而不必把全部信息都塞进工作记忆。
- 现代软件工程的瓶颈，正从最初的"生成代码"逐步转向"代码评审、架构设计"，以及管理复杂、由 agent 辅助的工作流状态所带来的认知成本。
- 依赖 AI 去生成 articles 或 communications 受到广泛批评：这往往适得其反。因为它会侵蚀 discourse 的质量，也暴露出对自己想法缺乏个人投入。
- 在 AI 时代，真正的速度（velocity）需要在 machine generation 的快与人类对理解的不可替代需求之间取得平衡。否则当"思考"被完全外包时，technical debt 与缺乏一致性仍然是高风险问题。
这次讨论反映出：在把人类的认知习惯适配到 AI-assisted development 的新现实时，人们面临着共同的困境。尽管不少人正尝试复杂的"manager-style"流程来并行化产出，但人们对这到底是提升了 productivity，还是只是用忙碌的幻觉遮住 burnout，普遍抱有强烈怀疑。那些相信未来的速度会解决当前挫败感的人，与那些认为真正的出路在于重拾刻意、缓慢且有意识的 human element 之间，分歧很大。最终，形成的共识是：tools 和 agents 应当用来增强专业判断，而不是取代 critical thinking 去避免积累失控的 technical 与 cognitive debt 。 • Productivity is currently being redefined by a transition phase where AI models are fast enough to require constant engagement but too slow to allow for true single-tasking, leading many to juggle multiple agentic sessions to avoid "idle" time.
• A growing consensus suggests that the "agent-heavy" workflow often mimics toxic, high-pressure environments like call centers, where success is measured by constant output rather than cognitive depth or sustainable output.
• Managing multiple autonomous agents creates a "context switching" tax that dilutes focus, leading to brittle, "tacked-on" code solutions that ultimately increase the maintenance burden for the human developer.
• Some developers have successfully adopted a "manager" scheduling approach, using Kanban boards and git-based issue tracking to treat themselves as supervisors who delegate tasks and review logic, rather than manual implementers.
• "AI brain fry" is identified as a real psychological toll, characterized by a loss of deep-work capacity and a subconscious tendency to mimic the staccato, overly formal writing style of LLMs, which many find dehumanizing and exhausting to consume.
• The most effective antidote to burnout may be a deliberate return to human-centric habits, including manual coding, reading books, and embracing idle time instead of reflexively starting a new parallel task.
• Sophisticated tooling for agent management is emerging as a necessity, with many developers building custom interfaces or using git-integrated project boards to provide the context needed to make informed decisions without holding everything in working memory.
• The bottleneck in modern software engineering is increasingly shifting from raw code generation to the cognitive cost of code review, design architecture, and managing the state of complex, agent-assisted workflows.
• Relying on AI to generate articles or communications is widely criticized as counterproductive, as it erodes the quality of discourse and signals a lack of personal investment in one's own ideas.
• True velocity in the age of AI requires balancing the speed of machine generation with the irreplaceable human requirement for understanding, as technical debt and lack of coherence remain significant risks when the "thinking" is fully delegated.
The discussion reflects a shared struggle to adapt human cognitive habits to the new realities of AI-assisted development. While many are experimenting with complex "manager-style" workflows to parallelize their output, there is a strong undercurrent of skepticism regarding whether this actually improves productivity or merely masks burnout with the illusion of busyness. A significant divide exists between those who believe future speed gains will solve current frustrations and those who argue that the real solution lies in reclaiming the human element of deliberate, slow, and intentional work. Ultimately, the consensus suggests that tools and agents should serve to enhance professional judgment rather than replace the critical thinking that prevents the accumulation of unmanageable technical and cognitive debt.