升职、跳槽、申请学位、争取项目机会时，最费劲的常常不是能力不够，而是临时想不起自己做过什么。
绩效评估时临时回忆一年做了什么，更新简历时翻遍邮件找数字，面试时举不出具体例子——这些不是能力问题，是信息管理问题。成长证据库就是为了解决这个：把你的量化成果、项目经历、他人评价、危机贡献、学习记录，长期放在一个随时可调用的地方。
不用一次写完，但要知道收什么。这 6 类覆盖了绝大多数职场场景需要的素材。
不需要复杂工具，不用一上来做数据库。先把结构搭起来，后面慢慢补内容。
每个文件不需要写成文章。一条一条记就行。比如量化成果：「报表手工时间减少 50%」——数字 + 背景 + 影响，一行搞定。重点不是一次写完，而是每次顺手补一点。
不要等到需要更新简历的时候才想起来整理。那时候你已经忘了大部分细节了。每次顺手补一条，30 秒的事。
AI 很适合从一堆材料里提取结构化信息。但哪些成果最值得讲、这次要强调哪条主线，还是要你自己来定。
如果你已经用 Claude Code 或 Codex 写过东西，可以更进一步——把证据库变成一个 AI 能读懂的项目。
具体做法：在文件夹根目录放一个说明文件（Claude Code 用 CLAUDE.md，Codex 用 AGENTS.md），告诉 AI 这个库是做什么的、有哪些规则、什么能改什么不能改。之后每次你打开 AI 编程工具，它就自动知道你的背景、你的主线、你的证据在哪。
不需要一开始就做进阶版。先建 5 个 Markdown 文件，开始记录。等积累到一定量了，再加 AI agent 管理。工具可以后面再升级，素材丢了就没了。
机会来的时候，能用的只有你已经记下来的东西。没记下来的，等于没发生过。成长证据库不是一个项目，是一个习惯。今天开始，以后每次顺手补一条就行。
When you apply for a promotion, change jobs, apply to school, or compete for a new project, the hardest part is often not ability. It is remembering what you already did.
Trying to remember a year of work during performance review, searching old emails for numbers before updating a resume, or blanking in an interview when asked for examples - these are not ability problems. They are information-management problems. A growth evidence bank solves this by keeping your quantified wins, project history, feedback, crisis contributions, and learning records in one place you can use later.
You do not need to complete everything at once. You only need to know what belongs in the bank. These six categories cover most career situations.
You do not need a complex app or database at the beginning. Build the structure first, then add material over time.
Each file does not need to become an essay. Write one line at a time. For example: "reduced manual reporting time by 50%" - number + context + impact. The goal is not to finish the archive today. The goal is to make adding evidence easy.
Do not wait until you need to update a resume. By then, you will have forgotten most of the details. Add one small entry when the event happens. It takes about 30 seconds.
AI is good at extracting structured information from messy material. But deciding which achievements matter most, which story fits the opportunity, and what to emphasize still belongs to you.
Please read the materials below and extract evidence for my growth evidence bank. Group items into quantified wins, feedback from others, project experience, crisis moments, and learning records. For each item, keep the original evidence, rewrite it in resume/SOP language, and mark what additional detail I should verify.
If you already use Claude Code or Codex, you can turn the evidence bank into a project that an AI agent understands.
Put an instruction file in the root folder: CLAUDE.md for Claude Code or AGENTS.md for Codex. Tell the agent what the bank is, what rules it should follow, and what it can or cannot change. Then each time you open the AI coding tool, it already knows your context, your main themes, and where the evidence lives.
You do not need the advanced version on day one. Start with five Markdown files and begin recording. Add an AI agent later when you have enough material. Tools can be upgraded later; lost evidence cannot.
When an opportunity arrives, you can only use the evidence you have already captured. What you did not record is effectively gone. A growth evidence bank is not a project. It is a habit. Start today, then add one small item whenever something happens.