[Cursor] Implement persistent context storage for AI tools

Something I don't see discussed enough: every insight cursor learns about your work — your stack, your conventions, your past decisions — disappears when the session ends. And the memory features that do exist? They're locked inside one platform. When everyone has access to the same AI, the differentiator isn't the code it generates. It's the context you bring. The hard-won lessons, the architecture decisions and WHY you made them, how your specific projects and clients actually work. That context is incredibly valuable — and you should own it. A few things worth considering: * Store your context as plain files you control. Not inside a vendor's system. Markdown, text files, whatever — something you can git version, grep, and move to any machine. * Keep it agent-agnostic. You'll switch tools. Your accumulated knowledge shouldn't have to switch with you. * Start now, even if it's messy. Capture decisions and lessons as you go. Structure comes later. The compounding effect of persistent context across sessions is significant. I've been working on an open source tool for this (Context Vault — local MCP server, plain markdown, works with Claude Code/Cursor/ any MCP client). But the tool is secondary to the idea. Fork it, build your own, use a notes folder — just start owning your context. I'll share the repo in the comments if anyone is interested.