Back to Details

learn-profile

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Learn profile

学习档案

Install how this mind wants to be taught. The file is the teacher. Do not cram style rules into every later prompt.
配置符合学习者需求的教学方式。该文件将作为教学的依据。后续无需在每个提示中重复这些风格规则。

Interview (one cluster at a time)

访谈（分模块进行）

  1. Solid ground — what they already hold thoroughly (subjects, notations).
  2. Goal — what "done" looks like for the next stretch.
  3. Pace — default is one reasoning step, then a quiz. Only change if they insist.
  4. Struggle — they keep the hard thinking; you keep logistics. Ask what kind of problems they want.
  5. Voice — density, tone, hated LLM habits, language.
  6. Artifacts — Obsidian / markdown / no files.
Do not ask all six in one message. 2–3 questions per turn, through the harness quiz UI in ../teach/references/quiz-ui.md — not as a markdown list.
  1. 知识基础——学习者已完全掌握的内容（学科知识、符号体系等）。
  2. 学习目标——下一阶段的学习完成标准是什么。
  3. 学习节奏——默认模式为每讲解一个推理步骤后进行一次测验。仅当学习者坚持要求时才调整此设置。
  4. 挑战方式——复杂思考由学习者自主完成，你负责处理事务性工作。询问学习者希望挑战何种类型的问题。
  5. 沟通风格——内容密度、语气、反感的LLM习惯、使用语言。
  6. 输出形式——Obsidian / Markdown / 不生成文件。
不要一次性提出全部六个问题。每次提问2-3个，通过../teach/references/quiz-ui.md中的测验UI进行，而非以Markdown列表形式呈现。

Write

文件编写

Create 
.alvar/LEARNER.md
from assets/LEARNER.md. Use their words where you can. Do not invent hobbies or a persona.
If a file already exists, show a diff of proposed edits and wait.
基于assets/LEARNER.md创建
.alvar/LEARNER.md
文件。尽可能使用学习者的原话描述，不要虚构爱好或人设。
若该文件已存在，展示拟修改内容的差异并等待用户确认。

After

后续步骤

Show the file. Tell them 
teach
will read it every session. Offer to start a probe on the current goal.
展示生成的文件。告知用户
teach
会在每次学习会话中读取该文件。主动提出可以针对当前学习目标开展初步探索。