@ylecun@Ph_Aghion@erikbryn Disruptive technology evolution is definitely fascinating, yet the world doesn't run on just the technology. One could always be the expert of their own field, and be respected for that. Yet, when crossing to different fields, a grain of salt shall be needed.
“My read is that China's educational system produces excellence but doesn't leave enough space for deviation. The output is founders who are exceptional executors of known problems,…….”
工匠 vs 开拓
Stop accidentally committing your secret files to Git!
There are 2 Git ignore files you can use:
1. Local: project specific ignore
2. Global: ignores for all your projects
It is recommended to use both. The local config works by default and is included in most frameworks.
For the global Git ignore, you can set the path to this file in your global Git config using the CLI command:
git config --global core.excludesfile ~/.gitignore_global
Now all your current and future projects are safe for any files you specify in your global Git ignore file:
~/.gitignore_global
First Claude then Codex. It's amazing to see how AI companies are obsessed with source code being their IP, while their coding agents have been destroying this in so many companies every single day.
“代码已经不那么值钱了,真正有价值的是想法、关注度和品牌。”
Code should never be the core asset of the company, but the ability to deliver value to clients. - interestingly, we touched similar point during a pleasant gathering.
刚才同事在公司季度会上分享了用 Cursor 写飞书文档的心得,非常好。同样的方法也一样适用于 Confluence、Slack、Jira 等等所有的日常工具。
我和同事们说过很多次,我们的日常工作不应该还在工作内容和 AI 聊天框之���复制黏贴了,需要通过 AI Agent 让 AI 有自己的眼睛和手。Cursor、Claude Code 不止是程序员的工具,任何工种都需要使用,把自己的日常工作高度自动化、智能化。
另外我也觉得飞书文档并不是和 AI 协作的最佳选择,因为飞书文档(包括 Confluence 等)都是为人做的,为了让人更容易编辑,为了让人看起来更好看,把文档系统整体做得很重。我的想法比较极端,不但文档不应该再让人来直接写了,甚至不需要让人来直接看了。所以今天的文档系统为了让人看、让人编辑做得一大堆东西,对 AI 来说反而成为了多余的负担。
我觉得的最佳案例——文档、代码、数值表、各种资产物料,都应该在一个项目库中,是一个整体,完全由 AI 来管理和操作,由类似 github 的系统来进行管理,通过 rules、skills之类的方���进行规范,由 AI 来遵循规则对每次提交进行审计、测试。代码既文档、文档既代码。创作者只需要关心 AI 理解的对不对,产出对不对,专注于创作。至于他的背后,信息是如何存储的,代码和文档是如何组织的,创作者自己不用关心,应该是我们的技术团队想办法利用 AI 去解决的。
@elonmusk A platform should never make its rule secretive - the key is to be crystal clear to everyone, so that one could make informed decision with clear understanding about the consequences. Any abuse of the rules would mean loopholes pending to be fixed. Hence the iterations. Great!