Hello you! 👋

On October 16, 2025, while the tech world was still digesting OpenAI’s Instant Checkout within ChatGPT, Anthropic quietly released something called “Skills.” No flashy demos. No performance charts comparing against GPT-5. Just a simple announcement about folders containing Markdown files and optional scripts. Within hours, Simon Willison—one of the most productive voices in AI development—called it “maybe a bigger deal than MCP.” Here’s what makes this strange: Claude has only 5% of ChatGPT’s user base. But it generates 40% of OpenAI’s revenue. ...

The Nature Editorial That Caught My Eye A recent Nature Reviews Bioengineering editorial titled “Writing is Thinking” made a compelling case for preserving human-written scientific papers in the age of large language models. The editorial argues that writing isn’t merely about reporting results—it’s a cognitive tool that helps us uncover new thoughts and ideas. As the editorial puts it: “Writing compels us to think—not in the chaotic, non-linear way our minds typically wander, but in a structured, intentional manner.” This resonated deeply with me, especially as someone who started this blog as a way to maintain “conscious effort” in my thinking. ...

When Dad Becomes Your Research Assistant My dad texted me a photo yesterday from his antique shop. “Found this old Cantonese children’s book from the 1920s,” he said. “Thought you might find it interesting.” I opened the image and immediately felt that familiar mix of nostalgia and confusion. I could mostly read it—traditional Chinese characters arranged in neat columns, reading right to left, top to bottom. But there were nuances I wasn’t catching, cultural references that felt just out of reach. ...

I just read an excellent deep dive from the Manus team about context engineering for AI agents, and it’s packed with practical insights that anyone building AI systems should know about. The Core Insight The Manus post1 opens with a striking observation: “Most agent failures are not model failures anymore, they are context failures.” We’ve moved past the era where model limitations were the primary bottleneck. Today’s models are remarkably capable—the challenge is giving them the right information at the right time. ...

I had that sinking feeling during Anthropic’s product demo last week. You know the one—when you’re supposed to be the expert in the room, the person who knows Claude inside and out, but suddenly everything feels foreign. The Anthropic team was walking through their enterprise offerings: Projects, Knowledge, the new Financial Services Analyst. Each feature seemed isolated, almost mundane when presented individually. Projects were just “workspaces to house user uploaded documents.” Knowledge was simply “parsing PDFs into markdown.” The Financial Services Analyst was just “MCP with enterprise support.” ...

Mark Zuckerberg is paying $100 million signing bonuses for AI researchers. Not $100,000—$100 million.1 And when you look at who Meta is hiring, a pattern emerges that reveals one of the most fascinating talent pipelines in modern technology. Seven of Meta’s eleven publicly listed Superintelligence Labs recruits graduated from prestigious Chinese universities before pursuing advanced degrees in the US.2 Shengjia Zhao, Jiahui Yu, Shuchao Bi, Hongyu Ren—these aren’t just individual success stories. They represent a systematic phenomenon that’s reshaping the global AI landscape. ...

I stumbled across Calvin French-Owen’s reflections on his time at OpenAI and found myself completely absorbed—not by the high-level insights about AI safety or AGI timelines, but by the operational details buried throughout. The kind of specifics that make you go “wait, they did what in 7 weeks?” There’s something deeply fascinating about getting a peek behind the curtain of a company operating at the technological frontier. Calvin’s post is full of these moments where you realize just how different building products looks when you have the resources, talent, and mandate to try audacious things from scratch. ...

Translation Note: This is a translation of the original Chinese blog post “写在 Kimi K2 发布之后：再也不仅仅是 ChatBot” by Justin Wong (bigeagle), published on July 13, 2025. All opinions expressed are those of the original author. Introduction Two days ago, the Kimi K2 we’d been working on for most of the year was finally released. After pulling an all-nighter before launch, I slept soundly for two days and finally have some time to write about my thoughts. ...

When Optimists Meet Realists Two posts crossed my timeline recently that perfectly captured the current moment in AI adoption. The first, from Yangyi, radiated the kind of infectious optimism you see from someone who just completed several projects with Claude Code: “After using Claude Code to complete several projects today, I have a strong feeling that the changes we envisioned in 2023 have now crossed a leveraged qualitative change moment… We can quickly use Claude Code’s Hook to complete long-process task planning and implementation. This means that 24-hour AI engineering teams will appear soon, and the entire process will accelerate exponentially.” ...

Note: This is a translation of the original Chinese article by grapeot, published on July 12, 2025. The original article can be found at: https://yage.ai/kimi-k2.html Recently, whenever a new large language model is released, we can always see a wave of evaluation reports. But there’s an interesting phenomenon: the vast majority of evaluations are invariably confined to an environment we’re most familiar with—the chat interface. This is actually a fundamental limitation. In 2025, when Agentic AI is all the rage, evaluating a model designed for Agentic capabilities within a chat interface is like using whiteboard coding to interview a development director. The feedback you get has almost no relation to its core capabilities. ...