Fields
as in field of vision, field of thought, and field notes
I forked Kevin Roose's open-source read-it-later app and rebuilt it around a different use case: visual research.
The original is a clean, text-first Pocket replacement — save articles, own your data, no subscription. It's great. But in UX/CX work, a huge chunk of what's worth saving isn't an article. It's a UI pattern, a screenshot, a campaign page, a product you want to reference later. Text-only tools don't serve that.
My version is built around a masonry card grid that renders different content types differently — articles, images, books, products, highlights, music, video, links, and notes each get their own layout treatment. Saved articles open in a full reading pane. Everything is searchable via full-text search, and you can filter by folder, tag, content type, or colour label.
Stash — The OG
The original interface was functional but optimised for text-based media.
The AI layer handles the tedious parts: GPT-4o-mini auto-tags saves on ingest from the title and content, and a separate edge function generates a TLDR summary for any article — useful for deciding whether something is worth a full read. For image-heavy research, there's semantic image matching to surface visually similar saves.
My Version
I created a visual-first interface that supports the way I think, work and take notes.
The most involved bit is the canvas — a freeform spatial board for thinking with your library rather than just browsing it. You can place saves as cards, add freehand text notes, and draw connections between them. Pan, zoom, drag — built entirely with pointer events and SVG, no third-party graph library.
Each saved item also supports rich notes with a WYSIWYG editor and split-pane Markdown preview.
Still vanilla JS, still self-hosted, still free. Built for how I actually work.
What is it?
The canvas — a freeform spatial board for thinking with your library.