@Rafal Glaz 5

@Ryan Doser I think the learning from this for me is the process and setup that could be repurposed for creating other 'teams' that do stuff. So I'm going to explore what else I could set up. I like to feedback and iteration elements.

@Rafal Glaz I think some of that is personal choice, time constraints, use cases. For me, the questions are, 1. Will it help and be better than me doing it? 2. Does it reduce friction? 3. Does it need to be the same every time, or does it benefit from judgment? That third one is where I draw the line. Often the answer is a bit of both in the same workflow, deterministic for the plumbing, AI for the thinking bits. And depending on your business, the "tasks that need creativity, different outcomes and reasoning" might not be the exception. They might be most of what you do.

@Henry Haevernick Henry, thanks for asking. Good questions, both of them. Yes, the system supports revenue directly. I run a workplace wellbeing training org, so everything I built that day (the campaign landing page, the CRM restructure, the automation monitoring) is part of running and selling actual services. Claude Code is not a side project or experiment. It is how I operate the business day to day without hiring staff. I am not currently selling the setup as a service or coaching others on it, though that may come later. Right now the value is in what it lets me deliver as a solo founder. On combining skills, not a stupid question at all. Yes, combining skills makes sense, but think of it as orchestration rather than one mega-skill. Each skill should do one thing well. Then you can reference multiple skills in a single task or build a workflow that calls them in sequence. I didn't, however, create all my skills as once as I wanted to drill down into what was going to be most useful. So I've built mine over time. The risk with one giant combined skill is that it becomes hard to maintain and debug. Keep them modular. Let the context or the prompt bring them together when needed. Once your skills are set up in the right place, Claude Code will often pick them up automatically based on what you are working on. You do not always need to call them explicitly. But keeping them well-named and clearly scoped helps, because it is pattern-matching on relevance rather than working from a formal trigger system. So good naming and clear structure in your skill files matters. One of @Ryan Doser's videos covers this really well https://www.youtube.com/watch?v=erkzROBDEFY. It's even more helpful to see the visual structure of this. I am still experimenting with this daily. Testing different skill combinations, orchestration approaches, and edge cases to understand both the limitations and the benefits I have not come across yet. I still can't help feel it's early days and I think there are many more surprises to come as we keep using it.

That was fun @Ryan Doser. Nice seeing a different format. Go Iowa 😀

@Ryan Doser The tea is great, I'll tell you about the video later 😉

Great chat again @Ryan Doser A few things that resonated: The efficiency argument is undersold. The discussion about the copy-paste walled garden nails it. Before Claude Code, I spent half my working day moving text between just about everything. It felt like a repetitive strain injury for my brain. That friction disappears when your AI can read and write files directly. The real shift though is what Adam describes as building a process rather than getting an output. Each session builds on the last. You're creating something persistent and transferable, not starting from scratch every time. His advice about starting small with research prompts and levelling up one session at a time is good. You don't need a full CMO orchestrator on day one. You need one useful workflow that saves you 20 minutes, then another, then another. One thing I'd add that Adam touches on but doesn't fully explore. The context management discipline he describes, being surgical about what's in your session, is a skill in itself. It's not technical. It's about knowing your own work well enough to direct the AI effectively. That's a real business skill.

Welcome @Jason Clark. Coming back to tech after 25 years is a big leap, but you've picked a good time and a good community for it. The Classroom is a great place to start if you haven't explored it yet. What kind of apps are you building?