I think I am a 3.5? What I need is more of the WHY do I need <new feature>. Like the Ralph-loop feature. Love it but I have been doing fine with out it. So do I need it or am I using it wrong. I suspect I'm using it wrong. Same with skills. I'm not sure I need it, but I think I do. I think I'm using Claude at a high level, but I'm pretty sure I can use it higher. What about auto-claude? What about using Claude with Ollama? What about GitTrees workflows. I need to ensure my Mac is basically coding so much that it's on fire. I'm really looking for a community that understands Claude code.
You're asking exactly the right question — "why do I need this?" is more important than "how do I use this." That's the gap I'm trying to fill. The Ralph-loop article comes this weekend, and I cover when you need it vs when you're overcomplicating things. Skills, auto-claude, GitTrees — all on the roadmap. And yes, I'm building a private community for exactly this kind of discussion. Thank you!
I’m in the 3-4 range—comfortable with prompting, systems, apis. Have a full stack web development and data and analytics background. Been bouncing between vibe coding and more traditional development trying to figure out how to leverage my technical background while purging old-style dev habits/attitudes. Need to ditch seven years of corporate conditioning and free up my mind and creativity!
Thank you for your comment! That corporate conditioning is real — I see it in a lot of experienced devs. The hardest part isn't learning Claude Code, it's letting go of the "I need to control every line" mindset. Your technical background is actually an advantage once you learn to leverage it differently. Upcoming issues will focus on workflows that let you think at a higher level while Claude handles the execution.
I’m starting from scratch. I’ve heard amazing things about Claude Code and Cowork. I’ve used Claude via Poe as a chatbot to help write python code as I’ve been working on an algorithmic trading program for a while. Lately, I’ve hit a plateau, as I need to give more complex instructions, which is beyond my current abilities. So this looks like the right next step.
That plateau is real — at some point, the chatbot back-and-forth can't hold the complexity of what you're building. You're not alone; most of us hit that same wall when working on a complex project. Claude Code is better because it sees your whole project, not just one prompt at a time. We're all learning fast here — I'll keep you updated, and feel free to reach out if you get stuck. Thanks!
I think I am in the 3 to 4 range as I have custom specialist agents using skills. I use ccs to orchestrate various models, save money and centralise mcp server config.
Thank you for your comment. You're ahead of most users. The orchestration layer you're describing is where the real power is. Upcoming issues on subagents and MCP will dig into advanced patterns for multi-model coordination. Would love to hear more about your CCS setup
I’m at 3.5 I’ve developed a few POC one which’s has turned into a solution our business. I want to learn more about setting better instructions and token management. My frustration is keeping it to keep track of tasks and avoid deviations
Thank you for the comment. Task tracking and avoiding deviations come down to how you structure your CLAUDE.md and define workflows. The short answer: break complex tasks into explicit steps Claude must follow sequentially, and use clear file boundaries so it doesn't wander into unrelated code. Hooks can also enforce checkpoints automatically — I'll cover that soon.
Thank you for your comment. That's exactly what this Masterclass Series is designed for. Follow along issue by issue — each one builds on the last. CLAUDE.md → Hooks → Subagents → MCP → Production deployment. You'll get there.
Thank you for your comment. You're right, most tutorials stop at the basics. The later issues in this series cover CI/CD integration, team deployment, and production patterns. That's the Level 5 content — I'm working on adding a real project to help learners go from idea to a shipped product.
I think I'm at Stage 5+. I have deployed apps from localhost to AWS using Terraform, Bedrock, from local CI/CD to Github Actions to AWS based CI/CD and beyond. I also reversed engineered Substack's internal API to create drafts on Substack straight from my local markdown articles. It automatically inserts CTA for "subscribe" "share" "paywall" etc, and it does it differently if it's paid, free, or founding member tier. It also creates social media posts at the same time and schedules the postings. So that's just one of the 80 or so projects I did with Claude Code since it came out, and I had over 1200 commits to Github last year. I don't need ClawdBot/MoltBot, because I've already built projects that do what that does already.
I'm probably about a 3. I really need to get Claude working more autonomously. I'm an financial services operations professional an I want to get Claude to do all the noddy work so I can focus on revenue driving decisions.
I am just getting started with Claude Code. I've seen impressive demos using it to build Agentforce agents within Salesforce, and I was impressed, but I haven't attempted it myself (nor is Salesforce development necessarily my target.)
I haven't built many projects from scratch using any agentic system. Some small useful CLI tools built using LM Studio is all so far. I believe I can get to a 2 or 3 fairly quickly. I'm interested in prototyping an app idea (not necessarily agentic in and of itself) and backing it with a set of agents. I hesitate to call my goal a PKM system, as I'm trying to make it more specific to software architecture work.
I'm also interested in understanding how longer-term software evolution and maintenance changes in this new working paradigm from what we "grew up with". Like Ariel Strong mentioned, I think there will be some corporate unlearning to be done.
I am somewhere in 3-4. I use MCP rarely, same with parallel jobs, didn’t use background tasks so far. I think i can optimize a lot and think i habe a high need to (feel) being in control.
I would put me to three. Programming and creating a brain for claude code is my next to do where I am working on. A brain where everything is defined.
* What is installed for Claude Code
* When something is not installed show Claude Code how to install it
* What projects I am working on and create a project based CLAUDE.md to give Claude Code a quick context about the structure and language of a project additional to the global CLAUDE.md of course.
* What skills, agents I am using and how to install them again when I am changing my macbook
* What MCPs I have installed for Claude Code, add the information in it's brain to know how the MCP is installed and what it needs
* Beside the Claude Code context the ai-brain has also the knowledge of which can be used only in claude code, gemini cli or which can be shared.
I am looking forward to leanr much more in your newsletter!
I think I am a 3.5? What I need is more of the WHY do I need <new feature>. Like the Ralph-loop feature. Love it but I have been doing fine with out it. So do I need it or am I using it wrong. I suspect I'm using it wrong. Same with skills. I'm not sure I need it, but I think I do. I think I'm using Claude at a high level, but I'm pretty sure I can use it higher. What about auto-claude? What about using Claude with Ollama? What about GitTrees workflows. I need to ensure my Mac is basically coding so much that it's on fire. I'm really looking for a community that understands Claude code.
You're asking exactly the right question — "why do I need this?" is more important than "how do I use this." That's the gap I'm trying to fill. The Ralph-loop article comes this weekend, and I cover when you need it vs when you're overcomplicating things. Skills, auto-claude, GitTrees — all on the roadmap. And yes, I'm building a private community for exactly this kind of discussion. Thank you!
I’m in the 3-4 range—comfortable with prompting, systems, apis. Have a full stack web development and data and analytics background. Been bouncing between vibe coding and more traditional development trying to figure out how to leverage my technical background while purging old-style dev habits/attitudes. Need to ditch seven years of corporate conditioning and free up my mind and creativity!
Thank you for your comment! That corporate conditioning is real — I see it in a lot of experienced devs. The hardest part isn't learning Claude Code, it's letting go of the "I need to control every line" mindset. Your technical background is actually an advantage once you learn to leverage it differently. Upcoming issues will focus on workflows that let you think at a higher level while Claude handles the execution.
Awesome. Sounds like just what I need!
I’m starting from scratch. I’ve heard amazing things about Claude Code and Cowork. I’ve used Claude via Poe as a chatbot to help write python code as I’ve been working on an algorithmic trading program for a while. Lately, I’ve hit a plateau, as I need to give more complex instructions, which is beyond my current abilities. So this looks like the right next step.
That plateau is real — at some point, the chatbot back-and-forth can't hold the complexity of what you're building. You're not alone; most of us hit that same wall when working on a complex project. Claude Code is better because it sees your whole project, not just one prompt at a time. We're all learning fast here — I'll keep you updated, and feel free to reach out if you get stuck. Thanks!
I am on levels 1-2. I want to learn more about the Claude’s capabilities. The biggest thing frustrating me about Claude is the
hallucinations.
Thank you for your comment. Hallucinations usually come from Claude guessing instead of knowing. Two fixes:
1) Give Claude a better context via CLAUDE.md so it doesn't guess about your project,
2) Use extended thinking for complex problems
Love this, I would say I’m between 1-2.
Thank you and welcome!
I think I am in the 3 to 4 range as I have custom specialist agents using skills. I use ccs to orchestrate various models, save money and centralise mcp server config.
Thank you for your comment. You're ahead of most users. The orchestration layer you're describing is where the real power is. Upcoming issues on subagents and MCP will dig into advanced patterns for multi-model coordination. Would love to hear more about your CCS setup
I’m at 3.5 I’ve developed a few POC one which’s has turned into a solution our business. I want to learn more about setting better instructions and token management. My frustration is keeping it to keep track of tasks and avoid deviations
Thank you for the comment. Task tracking and avoiding deviations come down to how you structure your CLAUDE.md and define workflows. The short answer: break complex tasks into explicit steps Claude must follow sequentially, and use clear file boundaries so it doesn't wander into unrelated code. Hooks can also enforce checkpoints automatically — I'll cover that soon.
I’m level 2 and want 5. Help me.
Thank you for your comment. That's exactly what this Masterclass Series is designed for. Follow along issue by issue — each one builds on the last. CLAUDE.md → Hooks → Subagents → MCP → Production deployment. You'll get there.
I’m excited by this journey
Between a 3 and 4. Not many tutorials focus on shipping the product
Or prod deployments: that can be pretty frustrating
Thank you for your comment. You're right, most tutorials stop at the basics. The later issues in this series cover CI/CD integration, team deployment, and production patterns. That's the Level 5 content — I'm working on adding a real project to help learners go from idea to a shipped product.
I think I'm at Stage 5+. I have deployed apps from localhost to AWS using Terraform, Bedrock, from local CI/CD to Github Actions to AWS based CI/CD and beyond. I also reversed engineered Substack's internal API to create drafts on Substack straight from my local markdown articles. It automatically inserts CTA for "subscribe" "share" "paywall" etc, and it does it differently if it's paid, free, or founding member tier. It also creates social media posts at the same time and schedules the postings. So that's just one of the 80 or so projects I did with Claude Code since it came out, and I had over 1200 commits to Github last year. I don't need ClawdBot/MoltBot, because I've already built projects that do what that does already.
I am in Level 1. Where can I find the full tutorial regarding Level 1?
I'm probably about a 3. I really need to get Claude working more autonomously. I'm an financial services operations professional an I want to get Claude to do all the noddy work so I can focus on revenue driving decisions.
I am just getting started with Claude Code. I've seen impressive demos using it to build Agentforce agents within Salesforce, and I was impressed, but I haven't attempted it myself (nor is Salesforce development necessarily my target.)
I haven't built many projects from scratch using any agentic system. Some small useful CLI tools built using LM Studio is all so far. I believe I can get to a 2 or 3 fairly quickly. I'm interested in prototyping an app idea (not necessarily agentic in and of itself) and backing it with a set of agents. I hesitate to call my goal a PKM system, as I'm trying to make it more specific to software architecture work.
I'm also interested in understanding how longer-term software evolution and maintenance changes in this new working paradigm from what we "grew up with". Like Ariel Strong mentioned, I think there will be some corporate unlearning to be done.
https://nandigamharikrishna.substack.com/p/the-claude-code-leak-how-512000-lines?r=8op1j&utm_campaign=post&utm_medium=web
I am somewhere in 3-4. I use MCP rarely, same with parallel jobs, didn’t use background tasks so far. I think i can optimize a lot and think i habe a high need to (feel) being in control.
I would put me to three. Programming and creating a brain for claude code is my next to do where I am working on. A brain where everything is defined.
* What is installed for Claude Code
* When something is not installed show Claude Code how to install it
* What projects I am working on and create a project based CLAUDE.md to give Claude Code a quick context about the structure and language of a project additional to the global CLAUDE.md of course.
* What skills, agents I am using and how to install them again when I am changing my macbook
* What MCPs I have installed for Claude Code, add the information in it's brain to know how the MCP is installed and what it needs
* Beside the Claude Code context the ai-brain has also the knowledge of which can be used only in claude code, gemini cli or which can be shared.
I am looking forward to leanr much more in your newsletter!