一次性软件与决策解析度
一切从一个十分钟的需求开始。在机器学习研发工作中我们用Labelbox来让人类标注数据。有一次我发现标注的数据质 …
Disposable Software and the Compressed Reality: AI Native Is Strategy Reconstruction
Disposable Software and Decision Resolution
It all started with a ten‑minute task. In machine learning R&D we use Labelbox for human data labeling. One time I noticed the labeled data quality was terrible, so I sent it back for relabeling. The next day the new results arrived, and …
超越DRY：AI原生软件工程的思考
半年前，我写了这篇文章。当时有两个论断：Claude Code虽然刚发布没人重视，但这种方便集成的的Agentic能力很重要；未来软件分发和开发的方式会有很大变化，Library as a …
Beyond DRY: Thoughts on AI-Native Software Engineering
Six months ago, I wrote this article. At the time, I made two assertions: although Claude Code wasn't getting much attention upon its release, its easily integrable agentic capabilities were crucial; and the future of software distribution and development would change significantly, with Library as a Service (LaaS) becoming an …
用一次摸鱼经历详解AI管理实战
虽然我重度使用 AI 工（摸）作（鱼）已经很久了，觉得AI倍增生产力是一件现实的事情，但最近摸鱼摸出的新境界还是让我觉得值得分享一下。
我是一个Applied Scientist。今 …