AI workflows

Most AI workflow content is screenshots of someone asking ChatGPT a question. That's not a workflow, that's a screenshot. The thing that actually changes is when an LLM can read what I've read. My saved articles, my notes, my bookmarked threads. Not the model's training cutoff, my actual context.

These posts are about that. How I build a personal knowledge base Claude can use. Why feeding raw HTML to an LLM wastes tokens and degrades the answers I get back. What an agentic loop looks like when it has access to a real reading library, not just whatever's in the prompt window. Where the line is today between Claude as a research partner and Claude as an autonomous agent. Some posts are workflow walkthroughs. Others are sharper opinions about where this is heading.

The common thread: AI gets useful when it has context, and the most underrated source of context is the stuff I already chose to read. Save exists to make that context portable. The workflows here assume you've either got a clipping habit or you're willing to build one, and you want to turn it into something that compounds month over month.