From Saved Links to Publishing Systems

Almost everyone who works with information has the same artefact: a pile of saved links that has outlived several machines and several tools. The links are good. The pile is not. It is searchable in theory and unreadable in practice, kept because letting it go feels like losing something, and never read because finding anything in it is too much work.

The first move is to turn the pile into a catalogue. A catalogue is smaller than a pile. It is a selection, organised with intent, where each entry says why it was kept. The cost of cataloguing is attention; the reward is that the result is actually usable.

The second move is to let the catalogue publish. Once the catalogue is the source of truth, a publisher is a thin layer over it: read the directory, render, write. The publisher does not own the content, which means the content survives the publisher. This is the whole shape of Citta reading from LinkCatalog.

The third move, which is the one this lab is for, is to let agents help with the cataloguing. Fetching, summarising, proposing tags, drafting the publishing note — these are exactly the kind of work agents can do in draft and a human can confirm. The pile becomes a catalogue; the catalogue becomes a publisher; the cataloguing becomes partly agentic. Each step is small. The compound effect is a different relationship with what you have saved.