Links should know why they exist

We've written a lot here about what MojoPad does. This is the first of a short series about why: the handful of beliefs that decide what we build, what we leave out, and what we're working on next. If you've ever wondered why MojoPad works the way it does, this is where it comes from.

It starts with something small: a link.

A link says that, not why

Type two square brackets, or in MojoPad just type a page's name, and two ideas are joined. It's one of the best things personal wikis ever did. Linking became so cheap that people started doing it while they thought, not after.

But look at what a link actually says. It says these two things are connected. It doesn't say how. Does this study support the claim on that page, or undercut it? Is it an example, a source, a later refinement? A month from now you won't remember, and the person you share your notes with never knew. A page with forty backlinks is a crowd with no name tags.

None of this is a new observation. People who designed hypertext systems decades before personal wikis existed worried about exactly this. Researchers in the 1980s gave links types. Open hypermedia systems kept links in a linkbase, a store of their own, so that the connections could carry meaning without cluttering the text. Ted Nelson wanted links that ran both ways and quotations that stayed attached to where they came from. Douglas Engelbart wanted every paragraph to have an address, so that you could point at exactly the thing you meant. Tinderbox has spent more than twenty years showing what a tool built on those ideas can feel like.

We stand on all of that, gladly. MojoPad's ideas aren't new. What we're trying to add is a way of getting them into everyday use.

Why the rich ideas didn't win

The honest question is why the popular tools left most of that richness behind. We think the answer is simple. Every extra decision at the moment of writing costs something, and people quite sensibly stop paying. If a link asks you to pick a type before you can carry on, you'll skip it, and soon nobody bothers. The fast tools won because they asked for nothing.

So we don't think the choice is between rigour and speed. We think the job is to make rigour cost almost nothing. That's the belief this whole series grows from, and in one line it reads:

Hypertext rigour at wiki speed. The machine can propose, you decide, and every link and every claim can show where it came from.

What that means in practice

Plain links stay plain. Write as you always have. A link with no reason attached is a perfectly good link, and nothing ever makes you add one.

A reason is one click away, when you want it. This is the part we're working on now. On a canvas, an arrow can already carry a label such as supports or contradicts, and MojoPad's local model can suggest one. We want the same for every link in a page: hover a link and mark it as supports, contradicts, extends, example or source. The reason is kept outside your text, so your prose stays exactly as you wrote it, and a page's backlinks can then sort themselves by reason. "What argues against this?" becomes a question your wiki can answer.

Every link should keep a receipt. This is on our list too, and not in the app yet: each link would remember who made it (you, auto-linking or the AI), when, and the sentence behind it. Links the machine makes would wait as suggested until you accept them, and a noisy batch could be dismissed at once. A wiki that links itself is wonderful until you can't tell your judgment from its guesses. Receipts keep the two apart.

Link to the sentence, write in pages. We don't think you should have to chop your thinking into atoms before you can point at it. Write whole pages, and when you need precision, go finer. In MojoPad today you can copy a link to a single passage of a page, and a quote link opens a PDF at the words you quoted. In MojoPad 1.100.0, out now, an answer about a paper can number its points: ask for links to the passages, or turn on Number every point to its passage in Settings (it's off until you do). Each number opens the paper at the passage it came from, so you can read a point and its source side by side.

Checking has to be nearly free

This matters most where AI comes in. AI is fluent, quick and tireless, and it will happily fill your notes with text that sounds right. Everyone agrees you should check what it writes. Almost nobody does, because checking takes longer than reading. If verification is a chore, it doesn't happen.

So we think the only workable answer is to make checking cost almost nothing. Ask MojoPad a question and the answer cites the pages it drew on. Ask it about a paper and it shows which passages it read, each a click away. And when a sentence in an answer reports figures that the passage behind its number doesn't show, that number is drawn more quietly, with the passage that does show them beside it, so the one link worth a second look stands out from the twenty that are fine. An assistant connected to your wiki can check its own quote links against the paper before you ever read the report. None of this asks you to trust the machine. It just makes it cheap not to.

A second brain stores. We want to help you make something.

There's a popular idea that a notes app is a place to keep things: capture everything, file it well, and retrieve it later. That's real and useful, and MojoPad does it. But anyone who has gathered a thousand highlights knows that gathering isn't understanding. The pile grows and the thinking doesn't.

We call what we're aiming at knowledge synthesis: making claims from sources, with a chain you can check. Not a summary floating free of where it came from, but a sentence you wrote, linked to the passage that supports it, linked to the paper that passage lives in. Links that know why they exist, receipts that say who made them and passages you can open in one click are all pieces of that chain.

We'll say plainly that not every note needs to end up somewhere. Some of the best notebooks in history are collections of fragments kept side by side for their own sake. A canvas full of loose ideas that haven't resolved yet is a perfectly good place to be. Synthesis is where we want MojoPad to help, not something it should demand.

Two promises underneath it all

Nothing should need learning before it's useful. The richness should be there when you reach for it and invisible when you don't. You shouldn't have to assemble your tool out of plugins before it can think with you.

Your notes must always be able to get back out. A tool that only takes things in becomes a vault you're afraid to leave. MojoPad keeps your work in files on your own disk, and its MCP server lets the AI tools you choose read, search and write your wiki, with you deciding what they may touch. Your thinking isn't ours to lock up.

What comes next

Each of these beliefs deserves more than a paragraph, and each one has shaped something specific in the app. Over the coming posts we'll take them one at a time: how a link can say why in one click, why checking the AI has to be nearly free, why you should be able to link to a sentence without writing in fragments, why your notes must always be able to leave, and a few more. For each one we'll show the feature it led to, and where we think we still fall short.

If you've been using MojoPad, you'll recognise some of this already. If you think we've got something wrong, we'd honestly like to hear it. The ideas we're building on came from decades of people arguing about how a tool for thought should work, and that argument is far from over.