your reaction after reading the post.
Sharing the workflow I landed on after a year of trying to read papers for school without losing track of what's where. If you've ever caught yourself thinking "I remember reading something about this... in some paper... a few weeks ago," this post is for you.
Three features changed how I read. The first one feels obvious in retrospect but most highlighters don't actually do it: **regional highlighting**. I can drag a box around a figure, table, or chart and it gets saved as a highlight just like a text highlight. For STEM papers especially this matters — half of what I'm trying to remember is sitting inside a graph or a diagram, and "highlight the caption" doesn't capture it. Now I come back to the actual visual content months later, and I can ask AI questions anchored to the boxed region too.
The second is **sections inside each PDF**. Before, I had a flat list of highlights per document, and a 200-page textbook would turn into 80 unstructured highlights I'd never want to revisit. Now every highlight goes into a section I name myself — "Chapter 4 — Photosynthesis", "Methods", "Results", whatever fits. The rename and reorder live in the highlights panel, so I rebuild a document's outline as I read. When I come back to study two weeks later, the structure is already there waiting.
The third is the one I underestimated until I had it: **cross-PDF AI query**. One Ask AI bar that searches across every PDF and note in my library at the same time. I can ask "what did I read about light reactions?" and it pulls the actual page excerpts from the three different papers that touched the topic, with `[p.N]` citations I can click to jump straight to the source. Before, I'd be tab-switching between PDFs and ChatGPT, copy-pasting passages and losing the source pointer. Now the source comes with the answer.
I built all of this into an app called Notie ([notie.app](https://notieapp.com)) because I got tired of bouncing between Notion + a separate highlighter + ChatGPT and losing track of which passage came from which paper. I'm the builder, so take that disclosure for what it is — but the underlying ideas (region-highlight figures, structure-as-you-read, ask AI across your whole library) are valuable in any setup that supports them. Sharing because the region-highlight + sections combo specifically isn't something I've seen written up much, and it solved a problem thtat I believe a lot of students suffer from.