Stimpunks Foundation

Mutual Aid and Human-Centered Learning for Neurodivergent and Disabled People

glossy obsidian stone on orange background

Filling the Garden: Obsidian, Web Clipper, and Ulysses

🗺️

Home » Field Guide » IT » Filling the Garden: Obsidian, Web Clipper, and Ulysses

A garden is only as good as what you put in it. This is the part nobody documents: how things actually get there — and how to capture a web page without quietly turning a copy into a fake original.

Start Your Own Knowledge Garden tells you to put your originals in one folder and never edit them. Fair enough. But it does not say how a web article gets into that folder at two in the morning when you are reading on your phone, and it assumes you are willing to open a terminal. None of what follows needs one.

Two apps do the whole job. Obsidian for capture, writing, and browsing. Ulysses if you would rather write there — it works on the same folders, and you can use both, which is what we do.

Clipping the web into your library

The Obsidian Web Clipper is a free browser extension — Chrome, Firefox, Safari, Edge, and the Chromium family — that saves a page into your vault as Markdown. It has a Highlighter for grabbing only the passages you care about, and it keeps everything local.

Point it at your library folder and it becomes your intake pipe. One click and a page you would otherwise have lost to a dead link is on your own disk, in a format you will still be able to read in ten years.

First, the honest part: a clipping is not an original

Our rule for the library is that it holds originals, unaltered. A clipping is not one. It is a conversion — somebody’s HTML, run through a readability parser, rendered as Markdown. Things get dropped. Layout, embeds, and anything rendered by JavaScript may not survive.

So why put clippings in the library at all? Because for a web page there is no original you can keep. The page is on somebody else’s server and it can change or vanish tomorrow. The clipping is the artifact of record — it is the only copy you will have. That is a good reason to keep it and a bad reason to pretend it is something it is not.

Which makes the frontmatter non-negotiable. A clipping with no source URL and no clip date is worse than no clipping: it is a wall of text that looks like a source, cites nothing, and cannot be checked. Six months later you will not remember where it came from, and you will either not cite it or cite it wrongly.

A template that keeps its provenance

Web Clipper templates use variables to fill in page data. These are the ones worth having, and they are exactly the ones that make a clipping citable:

---
title: "{{title}}"
author: "{{author}}"
site: "{{site}}"
url: "{{url}}"
published: "{{published}}"
clipped: "{{date}}"
kind: web-clipping
---
> Clipped from {{url}} on {{date}}. This is a converted copy, not the original
> page, and not our writing. Cite the URL above.
{{content}}

Other variables you may want: {{domain}}, {{description}}, {{selection}} for a selection-only clip, and {{highlights}} for the passages you marked with the Highlighter.

Templates can also fire automatically. Template triggers pick a template based on the current URL — a simple “starts with” match, a regular expression between forward slashes, or a schema: match against structured data. The first matching template wins, and the one at the top of the list is the fallback. So you can have one template for journal articles, another for newsletters, and never think about it again.

Set the destination folder and note name in the template’s own settings. Two things to get right there:

  • Keep the library flat. Ours has no subfolders on purpose — all the sorting happens in the index, where changing your mind is free. Point every clipping template at the same folder.
  • Use the title as the filename. If your index is generated by keyword-matching filenames, the filename is the shelf label. {{title}} gets a clipping onto the right shelf by itself; a timestamp or a slug lands it in the Unidentified pile forever.

One thing worth knowing about where clippings end up: in our setup the library is deliberately not in git — it is over a gigabyte of binaries. Clippings there are covered by Time Machine and the team mirror, not by version control. If you want your clippings versioned, either keep them small and text-only, or put them somewhere your repository tracks.

Writing your notes

Open the garden as an Obsidian vault and your notes folder is just a folder you write in. Two settings make it painless: set new note location to a folder, and point it at your notes inbox. Every new note then lands in the same place instead of scattering wherever you happened to be.

Everything is Markdown in ordinary folders, which is the whole point. Obsidian is a good window onto the files, not a place your writing is trapped. Delete the app tomorrow and every note is still there and still readable.

Keep the two folders honest. The library is for things you did not write and must not edit. Notes are for your own thinking, and you should edit them constantly. When a note stops being a draft and becomes something you want to keep as a source, that is a deliberate move you make on purpose — not something that happens by drifting.

If you would rather use Ulysses

Ulysses reads ordinary folders through its external folder support, so it can manage the same notes without owning them. Nothing syncs, nothing is imported, nothing is converted — it is an editor pointed at files on disk, and it notices when something else changes them.

That means Obsidian and Ulysses can both be open on the same garden, which is how we work: Ulysses for long-form writing, Obsidian for capture and for seeing the shape of things. Setup is covered step by step in Using GitHub with Ulysses via Gitfox, which also gets your notes into version control without a terminal.

Browsing what you have

This is where Obsidian earns its place, and it is the part that changes how the garden feels. Backlinks show every note that points at the one you are reading, so a source you cited once resurfaces on its own. Wiki links let you link to a note before writing it — which is the useful direction, because it records what you meant to think about.

And the graph view draws your garden as a picture. It is genuinely good for spotting two things you would not otherwise notice: clusters that turn out to be one idea under three names, and orphans — notes nothing links to, which is usually a sign you filed something and never came back.

An honest warning about the graph. On a garden of any size it is an undifferentiated hairball and people conclude it is a toy. It is not — it is unconfigured. Colour your groups (sources one colour, your own writing another, index pages a third) and it becomes a map instead of a knot. Until you do that, it is decoration.

What Obsidian will not do is find something by what it means. Its search is text matching, so it can only find words you thought to type. That is a different job, and it is step four of the starter guide.

Honest limits

  • Clippings degrade silently. A readability parser makes judgement calls, and a page that is mostly JavaScript may clip as almost nothing. Glance at what you saved before trusting that you saved it.
  • Paywalled and login-only pages will not clip properly, and clipping something you had to sign in to read does not make it yours to republish.
  • A clipping is a snapshot with a date on it. The live page can be corrected, retracted, or rewritten after you clip it. That is why clipped is in the frontmatter — and why you check the URL before quoting from an old clip.
  • One-click capture makes hoarding easy. The bottleneck was never saving things, it was reading them. A library of unread clippings is a to-do list wearing a costume.
  • The exact settings labels move. Both apps are actively developed, and the steps here describe what to accomplish rather than which button to press. If a name does not match, check the Web Clipper documentation — it is the source of truth, not us.