Andrej Karpathy just joined Anthropic, and the why behind it is way more interesting than the headline. In this video I break down how Karpathy's recent work on context engineering, the LLM Wiki, and /goal-style autonomous loops lines up almost perfectly with what Claude Code has already been shipping. Then I share three predictions for where Anthropic goes from here, including a possible app store for context and an education layer for packaging your own workflows.

So I'm sure you guys have heard the term "AI slop", and everyone sorta defines it differently. Maybe you think it's those TikToks of AI-generated fruits going on dates. Maybe it's infographics with misspelled words. Maybe it's something else entirely. But I want to talk about it in the context of communication. Internal, external, content you put out into the world. I write my LinkedIn posts with AI. My agent knows my business, how I write, how I speak. That's just how I work now. And there's nothing wrong with that. I think everyone should be using AI to write if it makes them more efficient. But this isn't a binary yes or no. It's a spectrum. Sometimes AI can draft and send automatically. Most of the time, I want it to just draft. Then I review. If someone sends me an email with em dashes everywhere, I don't actually care at all that they used AI. The fact that I can clearly tell it's AI-generated isn't the problem. What I do start asking is: → Did they proofread this? → Is this completely accurate? And subconsciously, I might start losing trust. Not just in the email but in the person who sent it. Our job here has changed from writer to reviewer. This quote has really stuck with me: "You can outsource your thinking, but you can never outsource your understanding." When your name is attached to the content, you take credit if it lands, as you should. But that also means you need to take accountability if it's incorrect. Taste and reviewing are becoming more important than ever. AI is super intelligent and powerful, but I don't want to see a world where we trust AI so much, that we stop reviewing things, and then the human on the other end of the content starts losing trust in us. That's why even though I write with AI, and people know that, I still try my best to disguise it and make it sound as "Nate" as possible. Check out the LinkedIn post I just wrote about this HERE

Claude Code just dropped Agent View, a new way to run and manage multiple sessions from a single terminal tab. In this video I walk through how it works, how to launch and navigate sessions, and the new /goal command for long running agents. If you've ever lost track of which terminal tab is doing what, this fixes that.

I builded this website with Claude, it's really fast and good!! The Idea is simple but it's good. It's Fully working all the links and all things inside it , I'm asking well showed. And in one prompt IT'S the important thing for me that it didn't taked that much time!! What's your experience to build websitesor apps with Claude!! How many of you tried to building websites about any of your work?? Is their any idea or something that I can improve in this website??

Printing Press lets you turn any tool into a CLI that AI agents can actually use efficiently. It gives you a catalog of ready to go CLIs and also helps you turn almost anything into a CLI in minutes. If you've ever watched MCPs eat your tokens for breakfast, this video shows why CLIs are the better path for agent setups.