I started using Roam and as a proper geek, dug through the data it sends back and forth about me and my notes in the browser console. It was doing access logs and some random day I saw some random dude’s name in the access log for my notes. I reached out to ask. They told me he was a new employee. I saw no reason to save personal notes and ideas on a platform where any employee can enjoy them. Thereafter I took my notes to tools i wrote myself. Very enlightening to the incentives for building such tools.

Wow, that's very icky.

Roam has always felt like a bit of a chore -- while it's easy enough to set up backlinks, having to do that one step has always been like a waste of time to me. This is the kind of thing that imo an agentic workflow could do for you:

- Just start typing

- Let the LLM analyze what you're typing, given the RAG database of everything else you've added, and be able to make those kinds of correlations quickly.

- One-button approve the backlinks that it's suggesting (or even go Cursor-style yolo mode for your backlinks).

Then, have a periodic process do some kind of directed analysis; are you keeping a journal, and want to make sure that you're writing enough in your journal? Are you talking about the same subjects over and over again? Should you mix things up? Things like that would be perfect for an LLM to make suggestions about. I don't know if Roam is thinking of doing this or not.

But... backlinks are fully automated. If you just make forward-links that you'd normally do in the course of writing.

You're thinking of an optional step of adding extra links "just because", but IMO that's as a learning process in the beginning when you're not used to adding any forward-links whatsoever.

Yes.

IMO the 3 table-stakes features for a notetaking app in 2025 are AI-powered search (including a question-answering capability), showing related / recommended notes (via RAG), and automated clustering (K Means + LLM) to maintain a category hierarchy.

I think this might be the most exciting use-case of LLM's I've seen suggested here. I've struggled with exactly this problem with note-taking and personal knowledge-bases.

I'm not a huge fan of bringing AI into everything but as I was reading the blog post, it did feel like this is the place where it belongs.

I'd love to have this but only if it runs entirely on my own machine or on a server I own. Uploading all my notes to somebody else's cloud is a nonstarter.

Can Ollama do this yet?

> But there’s one main reason that I don’t use it anymore: when I write my notes the thought, ‘Where am I going to put this?’ plagues me every time. It’s a direct and immediate pain. And it sometimes gets in the way of me even taking notes at all. I have this sensation many times a day and it’s deeply uncomfortable.

I had a similar problem when designing my personal management system last decade [1]. Every system you use, you have to stick to in order to get results. Sticking to a system can be emotionally draining to the point where you give up.

IMHO, that sense of emotional drain you get with fancy note-taking systems is tapping into something true. Only a small fraction of what we think we need to remember actually matters and will benefit from so much care to structuring it. The rest is a waste and a drain on our limited cognitive resources.

My solution is to initially write in a designated place that allows for less structure. In the to-do system, the main doc has a "landing zone" for action items to be quickly jotted down, then structured and organized later. In the project system, I'll have a "dump" file where I dump project thoughts that I'm not sure are important. I trust that if the ideas I jot down are actually important, the structure they deserve will come to me later.

Is that trust always right? Maybe not 100% of the time, but it seems like a more useful heuristic than "everything I put into this system needs lots of structure I don't feel like providing, so I don't, and it makes me feel like a failure".

[1] https://open.substack.com/pub/renormalize/p/my-markdown-proj...

the problem with any categorization is having to choose one and exactly one category. that's why i prefer tagging. i don't need to choose a specific category, instead i add any tag that fits.

Roam got me so frustrated I spent 3 years and likely many more making an html/htmx ai first replacement. It’s still pretty raw, but it does everything I want and more. https://grugnotes.com

A post on Obsidian and a post on Roam today?

I don't know about you guys, but I'm an Obsidian lover and that's not gonna stop anytime soon. IMO the big problem about what this guy is saying can be boiled down to this:

>My most common behavior is to Actually Write the Notes. That’s why Roam needs to help me with the thought, ‘I don’t know where to put this.’ If it does that well, it makes the vast majority of my time spent in the app a breeze. If it does that poorly, it makes my experience so painful that I want to switch systems.

The lesson of Obsidian for me has been that organization is creativity. If what you want is to have an ideological maid that can organize all your thoughts for you, then you're gonna have a bad time with any note taking service (although I'm sure you can develop llm plugins to do this in a way that you personally enjoy now.) What's beneficial about these note apps is that they put this issue directly in front of your face. Either rise to meet it or go back to pretending like organization doesn't matter and avoid the responsibility of creativity.

Using Obsidian goes through stages much like a growing business. You start and you have personal relationships with all the notes so you can remember them, but once you get enough notes you realize it's too much to manage just using personal relationships and you need to start implementing a system. As you get better, your system changes, leaving a paper trail of notes with different systems. That's why the only thing that I think these note apps need is a deprecation system, but otherwise IMO they're perfect.

How can you organize stuff in obsidian? You have folders and search and that's it. I was thinking of switching from Word docs and putting my faith in backlinks to keep everything together, but now I'm not sure.

Roam’s product people sprinted to the 100m line then Notion’s product people ambled past them and went on to run a full marathon.

Roam took absurdly long to finish loading, long enough that I would lose my train of thought and get angry at the tool instead of writing down what I wanted to write down. I would write in Sublime Text while Roam loaded, then paste it in. Then wonder why I was bothering with Roam when I could’ve already saved a text file with my note, to a directory somewhere.

And do you think Roam’s team fixed the performance problems? Ha.

Logseq has been my go-to for a couple of years now, it's datalog-esque query language is great for automated page generation, and it's implicit "indirect" links are also really nice- the block-level note primative fits very neatly in my head as well.

In 2020 so many people recommended this to me in support of my writing and related research. But my laziness saved me.

Not surprised to see this. Whats interesting to me in all this is the misplaced faith in emergent structure.

Roam bet on the idea that if you link enough atomic notes, structure will self-organize.

Which is such a weird fantasy if you spend a few minutes thinking about it. Try writing code like that or building a company or just about anything else! Why should notetaking and archive development be any different

It's clear you need some sort of editorial hand to create something maintainable and future proof. Like zettelkasten had Luhmann’s obsessive discipline behind it. Evidently roam had um. enthusiasm and javascript?

and yeah, it’s telling that the comparison is to IDEs. Imagine an IDE that dumped every snippet you typed into a graph database and expected you to recompile coherence out of it by browsing links. thats what roam felt like after the honeymoon.

In general most of Roam's target should want to lean harder into opinionated workflows. there’s a reason tools like linear or notion are winning. they’re structured enough to relieve cognitive load, flexible enough to adapt. Roam tried to be emacs, but turns out most users don’t want to configure their own productivity dialect.

also, lol at the idea of "automated taxonomy". The entire knowledge management industry keeps rediscovering ontologies like they’re new. We are probably going to reinvent OWL at some point and give it a name like "neuroschema" or something

Aren't you describing (and Roam using) what is essentially brain mapping, which is a well-established technology based on how our memories actually work?

I'm not a fan of neurophysiology analogies because it veer into pseudoscience, but I'll play along.

Roam implemented static bidirectional links and called it associative memory. in reality, it's closer to mind-mapping software with backlinks. So without mechanisms for reinforcement (surfacing old notes intelligently), pruning (forgetting irrelevant junk), or plasticity (reorganizing in response to use), the system becomes a junkyard of half-formed thoughts.

Brains forget for a reason, roam doesn't

You can't delete things?

Really, I think the user in that case needs to be much more choosy about what they put in the database. It will save them time and greatly improve the signal-to-noise ration.

Deleting takes a lot of mental and emotional energy.

Going through all your notes regularly and asking “should I delete this? Does it bring me joy?”

I do not understand how people have enough things to put down for posterity where they need linking between different documents, rather than a simple hierarchy. I suppose we are very different humans.

Hierarchies vary based on the application and your point of view. For example, let's say you have an entry for beer bottles. Does that go under beer or under bottles, or under glass-making or Charlie's hobbies or something else?

And your perspective today might differ tomorrow or in a year or 20 years. Think about Wikipedia (and other wikis) - there is no hierarchy. You can start at any point and, in a sense, there's a hierarchy of pages with the starting point at the top.

I suppose my observation is that I have never seen anybody actually have enough of notes for long enough of time (cannot imagine needing any of mine in a few years) for this to be a problem. Those people likely exist but I have not met them.

Tana, Capacities, Bear, Lgseq all have backlinks and other stuff from Roam for years now or so, thanks to Roam IMO. I wish they were able to make some good money from this innovation but they moved too slow at some crucial moment.

OTOH the app that really won was Obsidian, due to flawless execution with the "local first" principle. Being closed source and "not listening too much to the community" weren't issues, they just focused and improved consistently.

> Tana, Capacities, Bear, Lgseq all have backlinks and other stuff from Roam for years now or so, thanks to Roam IMO. I wish they were able to make some good money from this innovation

I debunked this myth on the prior discussion in 2022. Backlinks were a well-known idea in the wiki community, to the point that they were part of WikiMatrix, and it's almost certainly the case that Roam copied the idea. TiddlyWiki had backlinks at least as early as 2006. https://news.ycombinator.com/item?id=30330835

> It turns out that I am rarely in a position, while writing or thinking, where I want to glance through lots of old notes as a way to figure out what to say or do. Mostly that feels like sifting through stale garbage.

IMHO you may need to produce much better quality knowledge. You are missing out.

A good KB - personal or shared - captures high-value knowledge and lets you pickup where you left off, years later, with little effort. That way you are always working with the best knowledge you've ever had. What defines high-value?

First, it's high-impact - it changes things in significant ways: Trivia about C++'s origins is unlikely to be worthwhile; something from an expert that changes your whole perspective about C++'s design and applications may be. Also it's accurate, high-quality knowledge; otherwise the impact will be much reduced or it may even be significant in the wrong way - for example, Wikipedia IMHO doesn't qualify (in other ways too), but a lit review by an expert can be priceless.

Second, it's hard to replace: 1) Discovery is unlikely: you are unlikely to think of or encounter it next time, at least not unless you revisit the issue in depth. 2) It's hard to find - even if you think of it, you won't be able to find it or recreate it easily. Maybe it's buried in a book you won't remember. For example, if you have info on operating systems, you need little about Windows, Android, etc. because you use them daily (hypothetically) and info is easily available. Insights on TempleOS might be better, or from that keynote by Dennis Ritchie that you attended.

By capturing the value, you get much greater ROI - a lifetime or career of it - from your knowledge work. That also incentivizes deep, high-value knowledge work.

[2022], earlier discussion at the time (109 points, 103 comments): https://news.ycombinator.com/item?id=30320977

I think people overcomplicate note-taking.

Make topic-specific folders for discrete topics (e.g. recipes). Anything generic, put it in a big diary file with daily entries. It's easy to scroll through my past few days of notes, and after a few days I don't really need to reference uncategorized miscellanea for the most part. If I do, I can usually find it with ctrl-F-style text search.

At the end of the day/week/whatever, feel free to #tag anything you think you'll need to come back to or copy it into a topic-specific file. I mostly don't do this though. People feel a need to retain this big body of knowledge from their notes, but I think most notes are disposable. It's easier to wait a little while before reviewing & then decide what's worth saving, which is typically not much.

I agree. For most people, having a zero-friction way of recording a short thought or idea and being able to search for it later is more important than creating a vast network of connections. Tagging is "good enough" for most needs.

I ended up building my own app for my notes and it turned into a chronological feed of short notes, like a social media feed. I just recently added linking between notes, but honestly, I've found that it's not essential. Just having a way to search by text and tag covers most of my needs. The chronological order also makes it easy to find stuff that I wrote recently or to filter by date.

Since there are no files or folders, there's also zero friction when it comes to recording something. I don't need to think "Where should this go?" or "Is there already a folder or a larger note this should be a part of?" I think that has honestly led me to just down more thoughts and ideas than if I was trying to maintain a strict structure to everything. (There are downsides to that, though, as it may mean I have more noise in my system, making it harder to find actual notes of value long term.)

When I tried to read this, I was sent through several redirects with a total of somewhere around 80MB or more of data downloaded, to end up at an otherwise blank "Enable JavaScript and cookies to continue" screen.

Why do people tolerate the WWW working like this?

The founder raised $8M on Wefunder and seems too have taken it and run. He hasn't provided any updates for a few years now.

Are there extensions that e.g. use NLP/LLMs/vectors to suggest potential links from elsewhere in ones KB? Could be a fairly straightforward plugin.

(I haven't used Roam personally and have no idea if it even has a plugin architecture or is extensible, but this reminds me a lot of some of the knowledge management work we're doing with corporates)

It seems like it would work really well for someone who practices zettelkasten. I spent some time trying to learn to manage my knowledge using it with Roam but it never quite clicked with my way of working.

Another one for the “not really using Roam any more” box

Roam seemed to depend too much on legends about a guy who was a hyper genius. Roam's implicit promise was that if you use the hyper genius's method then you'll become a hyper genius too.

I used to use Roam, but they move like molasses, no new features or fixes for weeks. F’ing on cruise 40 in a 60 zone. Also it didn’t give me much “connecting the dots”. I went back to simple Apple notes, save myself some time trying to squeeze value from the subscription