As a total beginner, one thing I'm trying to do is get Claude's help as much as possible to help me implement this training. So today I copied all the Clief Notes Foundation training into txt files and created a Claude project to help me action it all. So far I'm very happy! Still working through it to refine. But overall thought it would be helpful to share the txt files if there are other beginners wanting to get Claude's help as a coach through this process too. (1) I attached all the lessons as txt files to the project (2) I added these instructions to the project: You are helping me implement what's taught in Clief Notes. It's about AI folder architecture. All lessons have been attached as txt files. Remember I am a complete beginner when it comes to coding. You are helping me figure out how to implement this as it relates to my business. Never use emdashes. Never use "it's not this, it's that" style statements. If you're not sure, say so. Ask me questions until you're 95% sure you have enough context to get a good output. (3) And I used this prompt: Okay so I was working through this training but got confused in lesson 1.2 because I wasnt sure if the folder was supposed to be my business name as a whole, or if it should be a smaller project like 'Branding Overhaul' or if that is still too broad, and needs to be 'website redesign' or 'anchor youtube content'. Basically the job that I am trying to do is actually map out: * the folder architecture that's right for our business. With the goal of reducing token usage, so you Claude have enough information in each folder to do the specific tasks, but not extra where it's chewing through tokens. * the exact md files I need, and help formatting and creating them too * what context is going to be helpful in each folder so I can gather that data/examples for you The overall thing I'm looking for is to have a really well trained AI that deeply understands our business. Along with a decent setup for different specific tasks to be executed really well.

You built your first folder. Vote below, then drop a screenshot in the comments so we can see what you came up with.

I've just started going through the training. Downloaded Claude desktop and VS Code. I also want to setup Node.js I understand the instructions say only do this if you're using Claude Code. And I know I'm not a developer. The most coding experience I've ever had was coding my HTML Myspace design as a young teen haha. But I want to learn Claude Code slowly. I'm getting stuck on the download page which isn't a good sign haha. The two options I'm not understanding... Using: - nvm - fnm - docker With: - npm - yarn - pnpm I screenshot it and asked Claude. It told me to download the prebuilt one, which I have done. I'm just checking if that's fine to use the prebuilt package on Node.js? And also sharing the feedback because as a total beginner it's a stumbling block I came across :)

"Something I'm finding very useful recently: using LLMs to build personal knowledge bases for various topics of research interest. In this way, a large fraction of my recent token throughput is going less into manipulating code, and more into manipulating knowledge (stored as markdown and images). The latest LLMs are quite good at it. So:..." https://x.com/karpathy/status/2039805659525644595 This is the exact shift we talk about in Clief Notes. Stop using AI only for one-off outputs and start using it to build a compounding system. The repetitive tasks get handed to Savants, your tools run as an Orchestra, and you step into the Conductor role. That is how you get better decisions, cleaner knowledge flow, and more time for the work only humans should do.

So I'm just curious how everyone is handling security and review around some of these tools that are becoming available unbelievably quickly. I'm sure everyone here knows about all the drama around Open Claw and the fake project and all the mayhem that Created. @David Vogel just posted a pretty great tool in a Comment called "Claude Code Users, You're Wasting Tokens." When I reviewed it, it definitely looks amazing. It's a huge project and it's been around for a little while. Some of these viral tools concern me because of the possibility of just one errant line of code that would allow for the unseen prompt injection. When you happen to trigger it, who knows what it gives access to. I myself spend a lot of time as a consultant in the corporate implementation and also teach as an adjunct professor at a small college here locally. I just completed some SecAI+ "Teach the Teacher" training so this is also top of mind right now. They don't really have any specifics either other than you must do your due diligence so I'm just curious. At this time I'm pretty careful to avoid brand new tools and tools that have instant virality because it just seems like there hasn't been enough time yet for that to be validated. I haven't found a tool that I believe in yet to do it and I don't have the time, or in a lot of cases the knowledge, to do it myself. That's it. I love everybody's input. I feel like this is a critical topic in this space and it can be often overlooked.