JB.
← Back to projects

Threadglyph: AI-Assisted Learning

ToolIn progress2026

AI-assisted editing and learning with anchored comment threads and changes you review before accepting.

New to portfolio

Threadglyph is an AI-assisted editing and learning tool for personal knowledge systems. You highlight a passage, or an image, and start a comment thread on it, the same way you would in a shared document. In the thread you can ask an AI to explain something or suggest a rewrite. It can run on a local model, so nothing has to leave your computer. Suggested edits show up as a before-and-after comparison, and nothing changes until you accept it.

Ask about any passage

Highlight a sentence, start a thread, and mention @local. Before anything is sent, a preview shows exactly what will go to the model and where. With Local, that's Qwen 3 running on your own Mac. The answer lands in the thread, anchored to the passage, and the note itself never changes.

Recorded live with Qwen 3 (8B) on a MacBook. Waits sped up and marked.

Turn answers into flashcards

The learning features are in preview and off by default in the current pilot. In that preview, tag a highlight with @knowledge and it becomes a concept note, with the original sentence kept as evidence. Tag it with @card and it drafts flashcards from that concept, ready to send to Anki.

Preview feature, recorded live on the local model. Waits sped up and marked.

Comment on your own writing

You don't need the AI to use it. Threads work like comments in a shared document: reply, resolve, and search them later. They hold on to their words while you edit the text around them.

No AI involved. Typing sped up.

Threads on images

Every image in a note gets a Comment button. An image thread works like any other, and it sits in the same panel as the text threads, numbered in reading order.

No AI involved.

Threadglyph works inside Obsidian as a plugin.