Part 2 of AI nuggets from an actual AI coach, because evidently 10 minutes wasn't enough for everything front of mind after this client call.
In this one: why your Motherthreads should write your prompts so you never think about prompt engineering again, the chat limits nobody warns you about (chat length, upload caps, and why "picking up where we left off" is not as seamless as advertised), why you should NOT burn your important chat threads on heavy MCP work, and the C-suite trick: treat Claude as your strategic advisor and let the inbuilt AIs in your tech stack (Notion AI, Copilot) do the heavy lifting on Claude's instructions.
Plus the simplest token-saving habit nobody does: stop uploading PDFs of plain text. Markdown files for documents, CSVs for spreadsheets. Quicker for Claude to read, lighter on your usage, longer chats for you.
Btw, VERYYY impressed with Claude Fable 5 in the first 48 hours of use. More on that over the next week.
LLM Tuning 101 is on my YouTube (the full first session of my coaching, free). LLM Tuning 102 is coming: skills, Cowork, plugins, connectors, the lot.
👱♀️🤖⚡️
From my own experience working with AI tools, I've learned that writing effective prompts can truly transform the output quality and reduce the time spent refining results. The concept of letting "Motherthreads" handle prompt creation is a game-changer because it automates the prompt engineering process, saving you mental energy and ensuring consistent communication with AI models. One overlooked challenge is the chat engine limits many don’t discuss openly—especially chat length caps and upload constraints. This often leads to interrupted sessions where continuing "where you left off" isn't smooth. I learned to segment complex projects into smaller threads to avoid hitting these limits, which keeps AI interactions seamless. Also, I avoid using my main chat threads for heavy multi-step complex processing (MCP) tasks. Burning those threads can clutter the context and reduce future session effectiveness. Instead, I treat the AI assistant, such as Claude, like a C-suite strategic advisor who delegates operational work to integrations within my tech stack, like Notion AI and GitHub Copilot. This layered approach leverages each tool’s strengths while streamlining productivity. A major discovery was switching from uploading plain-text PDFs to markdown or CSV formats. Markdown files for documents and CSVs for spreadsheets not only reduce token consumption but also speed up AI processing — resulting in longer, more productive chat sessions without bloating usage costs. Finally, being impressed by Claude Fable 5’s advancements highlights how rapidly AI assistants evolve. Keeping up with version updates and tuning large language models (LLMs) based on your specific workflows can unlock even more productivity. Free LLM tuning sessions available on YouTube are a fantastic resource if you want to deepen your skills. In summary, embracing prompt automation, respecting chat limits, efficiently managing AI threads, and adopting lightweight file formats significantly enhance your AI experience and output quality. These practical tips have saved me hours and improved creative outcomes, and they can help anyone looking to make the most out of AI tools today.