This is the second of two posts on A Programming Paradigm for Spatiotemporal Composability by Yifan Shi, Wei Zhang, and Tianyi Cui. The first post showed the agent in DeepSeek Harness building and ...

On October 7, OpenAI released Intelligent UI in ChatGPT. Instead of always answering with text, ChatGPT can now answer with charts, buttons, forms, and small interactive tools. It reminded me of s...

Temporal is a durable execution framework. It recently added a developer-preview integration for LangChain Deep Agents — the package is temporalio.contrib.deepagents, and it still carries an “exper...

左：千层饼；右：芝麻酱花卷。 我有一个 Vibe Coding 布道师朋友，他见我就问：”古法编程的屎山也多了去了，你怎么专挑 vibe coding 黑？” 我：不一样。古法编程的屎山像千层饼，一层一层堆上去的，有地质结构的层次特征。配上 git log，你还能考古，层与层之间的逻辑是找得到的。Vibe coding 的屎山像芝麻酱花卷，你 coding 一次，新屎老屎就从头到尾揉一遍...

I’ve been experimenting with on-device LLMs, and llama.cpp is the most practical way to run one on a phone: fast on CPU, GGUF quantization keeps memory usage sane, and it builds cleanly with the ND...

After reading the debate about slowing down AI, I thought of two books I’d read before, as a military history enthusiast: Command and Control: Nuclear Weapons, the Damascus Accident, and the Il...

I published a small macOS CLI tool called recall. Here is what it is, what it does, and how to turn it on. What it is recall is a one-line command you drop at the end of anything that takes a whi...

There’s a lot of debate about whether — and how much — you should review the code your AI coding agent generates. Can you vibe-code this feature? Do you need to fully understand AI-generated code b...

When you build a service on top of Temporal, you quickly run into a problem that pure request/response systems never have to deal with: the person who kicked off the work is no longer in the room w...

You browse HuggingFace looking for a model to run locally, and the file list looks like this: DeepSeek-R1-Distill-Qwen-7B-Q4_K_M.gguf DeepSeek-R1-Distill-Qwen-7B-Q5_K_L.gguf DeepSeek-R1-Distill-Qw...