laybak

the Ken Liu story linked in the article is actually spot on. would recommend giving that a quick read

I laughed. underrated comment

you beat me lol

this reminds me of this other flavor of "program as weights" I saw few months ago, putting deterministic programs inside the weights of an LLM

https://www.percepta.ai/blog/constructing-llm-computer

quite a few people I know consume edibles as the main form of cannabis. but it seems underrepresented in all the studies I've seen so far. as several others have pointed out, you would think the act of smoking would itself have nontrivial effects on health

same. the older I get, the more I appreciate the "middle of the world" attitude in classical Chinese teachings. there's no single approach that's always / universally right. so it's wise to avoid going to the extreme in any direction

this seems like a chain of good practices. though I find it hard to stay disciplined about keeping commits well scoped and well described

Written a while ago but spot on

AI can hallucinate, but I've been misdiagnosed by humans so many times too...

I've heard this experience from quite a few folks before, but this is my first time hearing about a mandatory 8 months quarantine as a consequence... damn

how representative are Semgrep's benchmarks? everyone seems to have their own benchmark these days (guess it's good "content marketing") I'm honestly losing track

agreed. we should normalize this level of skepticism/scrutiny for all claims from big AI labs

yeah smells AI generated to me too

haha same! "# new tabs opened" is a poetic metric

wow didn't realize background noise can get so sophisticated haha

Nice! Has anyone figured out a good way to export the sounds and the MIDI?

That's the idea! :)

Author here. Wow this got a lot more attention than I had expected. Thank you all for sharing your thoughts, including critical ones.

The point about not being superficial/artificial and instead "just being yourself" sounds fair. To clarify, the copying/imitating I was referring to was much closer to the "imitate, then innovate" sense. That is, even if you make the best attempt to copy someone you admire, you end up with a mutated copy that is uniquely yours (because of your biases and life experience).

And to the point about it being shallow and simplistic — I agree, and the brevity was intentional. I treat this a mental model that is not universally correct, but is sometimes useful. To make sense of the world, I would use an ensemble of these models.

And I love the quotes in the comments!

Wow I couldn't find a single comment about the app itself. And a significant portion of the comments are just so negative...

Wanted to point out that "playing dumb" is a really effective technique to get the interviewee to share their knowledge, and in a natural way.

And I have nothing but love for anyone who has the courage to have long-form discussions about technical, philosophical , and other intellectual matters in public.

I didn't get the last reference, but well done!

Hi HN, I built Rumin as a suite of knowledge management tools to for my own needs. It includes a visual canvas and a rich text editor for linking notes.

After releasing the web clipper earlier[1], this is the remaining frontend code, for the getrumin.com web app.

I know quite a few of us are here interested in knowledge management. Hope this code can help with your projects.

Cheers!

[1] https://news.ycombinator.com/item?id=24461070

interesting project! I like how all the information and context can be organized around a task

haha fair point. what do you think would be a better title? (or what made the title a bad one)

Hi HN,

Creator of Rumin here. It is not a polished product, but I think the community would find the concept interesting: https://getrumin.com/

Rumin is a combination of free-form drawing (create page blocks anywhere on the Canvas) and a flexible knowledge representation (undirected connections, hierarchical collections, and backlinks).

I am going for a free-form and spatial medium like paper, with the ease of edit and retrieval of a digital format. The way I currently use Rumin looks something like this:

- Always leave one Canvas tab open, and use it as my "mission control": Brain dump, random thoughts, and journal entries go here. And I use a different Canvas for each thinking context or topic.

- Start a new Canvas for each research, brainstorming, or learning session.

- Hide the pages from the Canvas whenever I'm done to declutter. They can be re-inserted (via search, along with the connected blocks) in any Canvas.

- Use the search bar, "related content" section, and the "fetch suggestions" to retrieve existing thoughts and reuse my previous work

It also comes with an open source web clipper to capture content from web resources: https://github.com/jhlyeung/rumin-web-clipper

And I plan on open sourcing more of the product in the coming weeks.

Would love to hear your thoughts, workflows, and use cases.

thanks for checking it out! you can swap out this API call here: https://github.com/jhlyeung/rumin-web-clipper/blob/master/ch...

sorry the code is a bit of a mess, I'm working on cleaning up the entire code base and documenting things better

totally! I see these tactics as observations about human behaviours and ways to navigate them. the patterns generalize across different domains and are useful for understanding the world

author here, this comment already made my day :) feel free to take the code and run with it, and let me know if you have any suggestions/questions

Thanks, kapnobatairza! working on cleaning up the product before opening up, will reach out to ask you more about your use cases :)

thanks neovive! Yeah Notion's web clipping and table storage approach is quite elegant. It only gets a bit clumsy when we get to the "power user" (not very common) use cases

Thanks Lucas! I was a Notion web clipper user as well. For me it worked for the most basic use case of saving a page into a table. But for me these use cases kept coming up: - An idea belongs to multiple collections, as opposed to a single "table" - There are usually properties/metadata I want to save (e.g. YouTube channel information), which would take multiple copy-and-pastes back and forth each time - Bi-directional linking of captured content - I wanted full-control over the captured data, for more advanced queries/filtering and it's sad that web clippers tend to be one of these "table stakes" features that companies build a basic version for, and not invest further in.

Quick answer for "Why should I use Rumin?" is: "Perhaps you shouldn't yet, but let's stay in touch and I'd love to hear about your use cases and other ideas."

The current version of Rumin is very rough, and there's an overwhelming list of improvements to make. This is one reason why I closed it for sign ups for now. But in the meantime, I feel there's a lot the community can do even with just the web capture component being open source.

Regarding your concern about data loss. I intend to open source more and more parts of the platform, and somehow figure out a model to make the development sustainable.

Thanks again for checking it out!

ohh thanks for the heads up! will look into this. Sorry, this is my first time sharing an open source project.

EDIT: I've added an MIT license to the code. Thanks again for pointing out