• https://github.com/clice-project/clice language server for C/CPP LSP
• write concise essays, less than 500 words
• https://macos-defaults.com macOS defaults
• <C-r>= on insert mode, will insert calculation result to buffer

• FOFA/Shodan/hunter.qianxin etc search engine can be used to find unexpected exposed services, e.g. port:9200 for ES
• summarize your tool in a markdown table with availability, use this prompt to check ChatGPT’s levels, and re-enable some skills
• :g/foo/s/bar/baz/ in vim can replace bar with baz in the line contains foo
• :v same as :g!, inverse searching. :v/foo/d will delete lines not contains foo

• g? in vim will encode the selected TEXT in ROT13
• :let @a = @a . '@a', append macro call to make the macro recursive, run to the end of the file
• https://instances.vantage.sh/, list and compare instances of AWS EC2/RDS etc
• fc -W will save history to ~/.zsh_history. DO NOT RUN history -c THAT WILL CLEAR THE HISTORY

The Python standard library offers many module-script utilities that can be used as command-line tools with the python -m syntax.

• python -m webbrowser https://pym.dev/p open URL in default browser
• python -m antigravity open https://xkcd.com/353/
• python -m this display the Zen of Python (PEP 20)
• python -m __hello__ print Hello world!
• python -m http.server start a simple web server
• python -m json.tool validate and pretty-print JSON data from stdin or file
• python -m calendar display a command-line calendar
• python -m uuid generate an UUID string
• python -m sqlite3 launch sqlite3 shell
• python -m zipfile/gzip/tarfile like zip/unzip or gzip/gunzip or tar CLI
• python -m base64 like base64 CLI
• python -m ftplib like ftp CLI
• python -m pip run pip to install third-party packages
• python -m venv create an virtual environment
• python -m pydoc show documentation for given string
• python -m pdb run Python Debugger
• python -m unittest run unittest tests from modules or file
• python -m timeit time a Python expression
• python -m site display current Python environment info like sys.path
• python -m platform display OS information
• python -m encodings.rot_13 encode/decode text by ROT13 cipher

More utilities and detailed explanation Python’s many command-line utilities

I’ve known @sindresorhus for a long time, used some of his Apps, and I’m a big fan of his work. Here are some of his Apps that I use:

Velja

Even though I prefer using Safari as my default browser, there are certain cases where I need to open a link like Google Docs in Arc because Safari is logged in with my personal account, while Arc is with my work account. In cases like this, I use Velja:

1. Set Velja as the default browser and Safari as the primary one
2. Use Arc as the alternative browser and click on links while pressing Fn for them to open in Arc
3. Set rules for specific URLs to open in Arc, with URL matching
4. Velja’s Safari extension also supports to open the current webpage in Arc

Velja is More Than a Browser Picker

Hyperduck

I need to send links from my iPhone to Mac for further reading, I have tried several methods:

1. Tools like Pocket/Instapaper, which require opening the app on Mac to use
2. Safari Reading List/iCloud Tabs, which I have been using for a long time before Hyperduck, but sometimes it does not sync well

Hyperduck does this job perfectly, shares link from Send to Mac, opens on Mac directly, and I don’t have to worry about syncing.

Thank you @sindresorhus for these great Apps! 🎉

From 29/02/08, this site has been running for 16 years.

1. debounce 延迟生效，将间隔不超过设定时间的多次连续调用变成一次。如果在设定时间内连续两次调用，第一次调用会被取消，如果设定时间内没有再次调用，则生效执行
2. throttle 节流阀，确保函数被多次连续调用时，在设定时间内最多只执行一次
3. debounce 和 throttle 都可以用于降低事件处理函数的调用频率，以提高性能
4. 在连续快速输入时，debounce 会等待最后一次输入后才执行，autocomplete 场景很有用
5. throttle 会规律、稳定的执行，但是会有一定的延迟

• https://reorx.com/blog/debounce-and-throttle/
• https://kettanaito.com/blog/debounce-vs-throttle

After changed ~/.ssh/config, Ace still can’t connect to MySQL, we need to config Ace to grant access to .ssh files:

devv.ai 是如何构建高效的 RAG 系统的

1. https://twitter.com/Tisoga/status/1731478506465636749
2. https://twitter.com/Tisoga/status/1736544319199478175

如何让 LLM 使用外部知识库进行生成？之前的做法是在增加新的知识库后 fine-tuning，缺点是：每次更新知识都要重新 fine-tuning，带来巨大的训练成本。新的方案是 RAG，Retrieval Augmented Generation（检索增强生成），通过 prompt 的方式把新知识给到 LLM。三部分：

1. LLM，GPT 或者开源的 LLaMA
2. 固定不变的外部知识集合
3. 当前场景下需要的外部知识

Notes:

• 外部知识库的存储，通过 OpenAI embedding 模型把知识数据向量化
• vector 向量数据库存储，Chroma、Pinecone、pgvector 等
• 优先做工原则：encoding 的时候做的越多，retrieve 的时候就能够更快更准
• 对数据做更多的细致处理，比如知识文档 chunk 分块，ranking 优化等
• 可以结合搜索引擎提高准确度
• 评估指标
  1. fluency，流畅性，生成的文本是否流畅连贯
  2. perceived utility，实用性，生成的内容是否有用
  3. citation recall，引文召回率，所生成的内容完全得到引文支持的比例
  4. citation precision，引文精度，引文中支持生成内容的比例

使用 LLM 的三种方式：Prompting, RAG, Fine-Tuning: RAG 用于扩展知识库，微调更多是关于改变结构（行为）而非知识。

I’ve set $CDPATH in zsh for quick directories switching. But this bring some issues with make or npm run:

/bin/sh: line 0: cd: src: No such file or directory

cd wants to change some directory but uses $CDPATH first to find. Add . to $CDPATH to fix this:

export CDPATH=.:$HOME/src