In this video, I’m basically trying to help you squeeze the maximum ROI out of a Claude subscription. Claude is genuinely strong right now—not just because the models are good, but because of the connectors and the different ways you can use it: the desktop app (Chat, Co-work, Code), the regular chat interface, and Claude Code in the terminal. I start with Chat because it’s the “second brain” use case I actually use daily—researching video ideas, arguing through architecture decisions, or just learning something new. If you want better output, my advice is simple: use the best model (I stick to Opus 4.6 Extended) and ask for deep explanations + diagrams when it makes sense. Then I show Co-work, which I think is most useful for non-technical folks or when you want Claude to operate on a folder and produce artifacts (like summaries or CSVs). After that we get into Claude Code the way I prefer it: in the terminal. I cover the practical workflow stuff—plan mode (Shift+Tab), effort levels, “ultra think,” the “by the way” branch for quick side questions, bash mode with “!”, checking context, adding multiple directories for cross-project reasoning, remote control via a session URL, and exiting cleanly so you can resume later. I also call out usage management (weekly limits) and the 2x usage windows, plus the reality that “dangerously skip permissions” is powerful but you need to understand the risk.