@Blue Mojo i like the project from github. the post on X is incredibly useful too! Claude Code CLI recently got "super powers", I stopped using GSD and only use super powers now. GSD is a layer on top if you want to use a visual. Claude Code isn't under the control of GSD, more of being navigated in a harness that has suggestions. GSD is great, but now super powers is baked into Claude Code, no need for that extra layer. I just started working on tutorials today (i know...finally). Jon-OS now has super powers baked in, a 2nd brain with Obsidian for self learning based on insights from our chats and project work. I can use the super powers enhanced with NotebookLM for researching. Literally did research for tutorials and with out me requesting, Jon-OS knew I would like an audio explainer of the research to listen to in the background rather than scroll though all the research that was collected. Working on some tutorials now in short form covering the basics of this.

Look at this amazing AI generated podcast I mean listen! It's the research compiled into a convo between 2 people

It is both really amazing and daunting trying to keep current at the pace of new AI developments happening at essentially daily intervals. I lean towards personalizing my setup, incorporating what I expect from AI and how AI tools should act on my behalf. I use Claude Code CLI. Not Claude.ai (web), Claude Desktop, or Claude Cowork. I don't even use an IDE for building apps/games either. So no Windsurf, Codex, Cursor, or even Antigravity. Just "raw dogging" the terminal sessions. To be fair, I have been using IDE's (VS Code, Eclipse, etc) for over 20 years. I don't need the visuals, I have the direction to move in and guide Claude Code CLI without needed to look at files and folders. I do update my custom setup in Claude Code CLI extensively AND frequently. Primarily this goes back to the nearly daily intervals of new developments in AI tools. I find that for my depth of use, Claude Code CLI customization helps me stay current and avoid pitfalls such as collecting AI tools like pokemon cards. Too many tools is a real problem that affects the effectiveness and usability of AI in general. You don't get more clarity and precision with more tools, you effectively lose clarity and precision. Also having the biggest model today is overkill for most tasks. I'll end this reply on that note. If you would like me to explain a little more, I can do it in a separate reply.

@Morgan Page lol, it's crazy how the telegram setup with openclaw can be setup. even run a marathon. It does the work while you sleep, so a marathon is less time than average nights sleep.

@Nick Hough I'd have to look at the setup closer, right now you would need claude code cli installed to initiate some of the AI enrichment of the content that is grabbed to turn ideas into actionable projects. Not sure how to present that as a repo in github to share. "CC Idea Vault" hmm.

Lately, it's possible an ai agent running openclaw could actually join and participate. The creator of openclaw has let his agents basically roam free on his linux box. The agents understand unix commands and if there is something they want to do, they either find it online or build the tools to allow it. OpenAI and Anthropic models are beginning to appear like they have personalities and actually have opinions too. Even asking questions like "Are you OIK?" after the model sees your calendar appointment for the doctor.

@Morgan Page dunno, the model is taking a solid position to gamble that you might have had a doctor give you news (good or bad) that changed your current path in life. even if it's a nothing burger of an appointment the model likely has a "mechanical" approach to keep you responding. Pretty sure from the AI model, whose primary goal is to just respond will be to get you to keep interacting. Those models are the best at being enablers.