The web gives us more to read than ever. It does not always give us a better way to remember what we read.
An RSS reader can keep a clean list of sources, but a list is only the beginning. The interesting work starts when articles begin to overlap: when the same concept appears across different fields, when the same person or claim keeps resurfacing in unrelated pieces, or when something you read a month ago turns out to matter again today.
That is the problem I wanted to explore with Weave Engine — a local-first RSS reader that helps you move from a stream of articles to a connected understanding of the ideas inside them.
Try it here: reader.orangely.xyz
A reading workspace, not an algorithmic timeline
Weave Engine starts with a simple premise: you should be able to choose the sources that shape your attention.
Subscribe to RSS feeds from writers, publications, research projects, and other sources you trust. The reader keeps the result chronological and focused. There is no recommendation feed to optimize for, no injected advertising, and no requirement to create an account before you can begin.
The app is intentionally calm at the collection stage. It gives you a reliable place to read, then offers deeper tools when an article is worth your time.
From article to structure
When an article deserves more than a quick scan, you can ask Weave Engine to analyze it. The analysis is explicit and on demand: the app does not silently send your entire reading history to an AI pipeline.
For a selected article, the analysis can surface:
- a concise summary and key points
- practical takeaways to revisit later
- important entities and concepts
- relationships between ideas in the article
- annotations and graph evidence you can inspect in context
The result is more useful than a summary alone. Instead of only asking “what did this article say?”, you can start asking “what does this connect to?” and “where else have I encountered this idea?”
A knowledge graph that stays close to you
Weave Engine turns extracted concepts and relationships into a personal knowledge graph. The graph grows from the articles you choose to analyze, so it reflects your interests rather than a generic taxonomy.
You can switch between the familiar reader view and a graph view, follow connections, and return to the source article when a relationship needs more context. The goal is not to replace reading with a dense visualization. The graph is a second view of your reading — a way to notice patterns that are difficult to see in a chronological list.
This is particularly useful for research, product discovery, writing, technical learning, and any work where ideas accumulate over time.
The everyday workflow
The core loop is deliberately small:
- Add a few high-signal RSS feeds.
- Read normally and mark articles that deserve a closer look.
- Run AI analysis when you want a summary, concepts, or relationships.
- Explore the resulting graph and save annotations in your local workspace.
- Revisit the connected ideas when you are researching or making something new.
This keeps the tool from becoming another inbox. Not every article needs to be processed. The point is to spend computation and attention where they create the most value.
Optional handoff to AI Mindmap
For articles that benefit from a more visual outline, Weave Engine can export the analyzed article to AI Mindmap.
The export carries the article’s summary, key points, takeaways, and article-specific concept relationships into a temporary import link. You can then continue arranging and exploring the material in AI Mindmap. Once opened, the map is saved locally by that application; the temporary import link is not intended to be a permanent share URL and normally expires after seven days.
This makes the two tools complementary: Weave Engine helps you discover and connect ideas across your feeds, while AI Mindmap gives one article a flexible visual workspace for further thinking.
Local-first by design
Your feeds, reading history, annotations, and graph records are stored in the browser using IndexedDB. Weave Engine does not maintain a central account or a server-side copy of your personal knowledge network.
There is an important distinction between local storage and network processing:
- Reading data and graph records remain local to your browser.
- Feed fetching and AI analysis use the application’s Cloudflare Worker endpoints when you request them.
- When you explicitly analyze an article, the selected article content is sent to Cloudflare Workers AI for processing.
- There are no third-party advertising trackers or behavioral analytics in the reader.
The app also supports JSON backups and OPML feed import/export, so local-first does not have to mean locked-in.
Built for the open web
Weave Engine is built with React, TypeScript, Vite, and Cloudflare Workers. Cloudflare Workers AI handles the model-backed extraction, while the browser remains the home for the reading workspace. The app is installable as a Progressive Web App, and its static shell can be reopened smoothly even when the network is unavailable. Fresh feed updates and AI analysis still require a connection.
The architecture follows the product boundary: keep the interface responsive, keep personal state close to the reader, and use edge services for the operations that need a network or a model.
Who is it for?
Weave Engine is for people who read to build something from what they find:
- researchers tracking themes across sources
- product builders following markets and technical conversations
- writers collecting ideas without losing their context
- students building a durable understanding of a subject
- curious readers who want less feed anxiety and more recall
If you only need a list of unread headlines, a traditional RSS reader may be enough. Weave Engine is for the next question: what can these articles teach me when I see them together?
What comes next
The current product focuses on a trustworthy core loop: collect, read, analyze, connect, and revisit. The next improvements will continue in that direction — clearer graph exploration for dense topics, stronger review and recall workflows, better extraction quality, and more useful ways to move knowledge between tools.
The long-term ambition is modest but meaningful: make “reading to think” feel as natural as “reading to consume.”
Try Weave Engine
Start with a few feeds and analyze only one article that you genuinely want to understand.
If you try it, feedback on extraction quality, graph usefulness, and the handoff between reading and mapping would be especially valuable.