Structure > prompts
|
I'm currently on vacation and trying not to work... Obviously, that's going real well... Before I jumped on a 21 hour flight, I took a day to work on the structure of my AI systems, Skills, agent files, prompts, project briefs. The lot. Mainly because I noticed that every time I'd put Claude to work, I'd get different outputs. Sometimes it would be in my brand voice. Sometimes the details would be correct. I dug into the way I'd built everything and noticed that, over months of building with Claude, I had a tonne of duplicates for things like ToV, plans, and more. Every time Claude went looking for context, it would find bits in different places which confused it. So I stripped back everything and worked on building a more robust file hierarchy within my agent systems. Something that made it super simple for me to know where to put key files, and for Claude to find them. Took like an afternoon to fully organise this stuff. But, since doing so, Claude's been better from cold starts and knows what we should be working on and the right tone of voice. All cause it has a better "map". Made me think that there's probably a lot of folk out there who have similar issues with Claude, and a simple system architecture could help. So, I'm building that as a freebie soon to be released. The basic folder architecture to get you started in the right way with Claude so you don't waste time. AND, a couple of the universal rules I use across all my brands that strip out the generic AI-tells and generic BS we all hate. Idea is you grab this, drop it into your system, and day one your AI sounds less like AI and more like your business because it knows how to think about you and your brand. Won't be out for a couple of weeks. I want to record a proper walkthrough video to go with it and I can't do that from a beach. More on this soon - I'm off to the beach now. Pete "Sunday folder reorg" Boyle |