I think all this discussion around Open-source AI is a total distraction from the elephants in the room. Let's list what you need to run/play around with something like Llama:
1. Software: this is all Pytorch/HF, so completely open-source. This is total parity between what corporates have and what the public has.
2. Model weights: Meta and a few other orgs release open models - as opposed to OpenAI's closed models. So, ok, we have something to work with.
3. Data: to actually do anything useful you need tons of data. This is beyond the reach of the ordinary man, setting aside the legality issues.
4. Hardware: GPUs, which are extremely expensive. Not just that, even if you have the top dollars, you have to go stand in a queue and wait for O(months), since mega-corporates have gotten there before you.
For Inference, you need 1,2 and 4. For training (or fine-tuning), you need all of these. With newer and larger models like the latest Llama, 4 is truly beyond the reach of ordinary entities.
This is NOTHING like open-source, where a random guy can edit/recompile/deploy software on a commodity computer. Wrt LLMs, Data/Hardware are in the equation, the playing field is complete stacked. This thread has a bunch of people discussing nuances of 1 and 2, but this bike-shedding only hides the basic point: Control of LLMs are for mega-corps, not for individuals.
But there is an insidiousness to Meta calling their software 'open source'. It feels as if they are riding on the coat tails of the term as if they are being altruistic, when in fact they are being no more altruistic than any large corporation that wants to capture market share via their financial muscle - which I suppose touches on your last point.
Thank you for this comment - you have succinctly captured something that I have been feeling but unable to express in words.
Stories/memes/narratives are the most easy/potent form of mental ingestion - so much so that I think humans cannot ingest facts or ideas at all, only stories. And this puts a collective responsibility on all of us to be very careful about the stories we create.
I strongly recommend everyone read Luhmann's original paper [1] on Zettelkasten.
What I got from it is that, the process is less about storing/retrieving information, but more about building a system that can surprise you - the way ideas emerge when you brainstorm with others. These others being people with the same level of knowledge as you - so, not the kind of information flow that happens between an expert and a beginner, but between peers who have access to the same kind of information, but simply look at things differently.
For some reason, every Zettelkasten system out there focuses on the mechanics, and in my opinion, not enabling the original intent of the approach. Not that I have a solution to the problem.
I used the Zettelkasten idea for a bit more than a year, and it was truly a joy to use. Unimportant idea get naturally forgotten, while important/relevant ones get reinforced and revisited a lot. Building the links feels like having a deep discussion with a knowledgeable friend, and, once in a while, you get that surprising insight when a note from a book about a completely different topic you forgot you read pops up and sheds a new light on your current reading.
But the main weakness of that system is what makes it so powerful: it is, by nature, extremely time consuming. It is so powerful because, in order to add the smallest note, you need to sift through dozens of them, amend them, thinks about connections, create notes for those connections... In a sense, it is the exact opposite of what the article describe: the system _prevents you_ to just jote a quick note and forget about it. I does make it very powerful as a "second brain", but I ended up finding that the joy and insights I got out were not worth the time investment. I guess this is a reason app builders do not emphasize the process: "get more out of your reading by spending a few more hours per day organizing your thoughts" is not super sexy.
I now see that system as relevant only to people whose _main_ focus is to collect information and extract new insights from them: philosophers, anthropologists, some types of sociologists... But if you just want to remember a blog post about weird CSS tricks, this is probably overkill.
> In a sense, it is the exact opposite of what the article describe: the system _prevents you_ to just jote a quick note and forget about it
Create daily notes (tied to the day you wrote them) or transient notes (tied to a concept) as scratch areas to quickly dump your thoughts onto the page.
Then return later when you have the time and focus to explicitly process your notes - either discarding daily/transient notes that are no longer interesting, or growing them into evergreen notes with sufficient detail and linking. Make sure it's easy to list all your daily and transient notes in an "inbox".
> For some reason, every Zettelkasten system out there focuses on the mechanics
Yes. The Zettelkasten system is not equivalent to an unstructured wiki. It's more than just docs with links. I'm no expert, but I think of Zettelkasten as describing a way to work with your collection of notes over time.
> For some reason, every Zettelkasten system out there focuses on the mechanics, and in my opinion, not enabling the original intent of the approach.
Probably because most people don't learn the intent. Most people don't even seem to understand what a Zettelkasten actually is, and take it for the name of the system. It's a very common problem with most of those systems and tools. GTD and Bullet Journals have the same problems. I think it's because people only learn the culminating result, the system, but not the way that lead to the system and tool, and reasons why the system and tools are the way they are. And then they start the cargo cult and praise the name.
Shameless plug. For those who use Orgmode, but think that Agenda is a bit too heavyweight to manage Todos, I created a Magit like interface to manage Todos (in an orgmode file): https://github.com/ChanderG/toodoo.el
Aside - this is a great example that shows how bad Google search has become. Seeing only the comments here, I tried to search for "pgpodman" not realizing that it's from this very article. Google only returns stuff about podman. Searching for "postgres pgpodman" only gets articles about pgAdmin.
DuckDuckGo gets this article as the first hit - top 4-5 links are all relevant.
The way I see it, in Emacs it's less an "extension" and more of "I have this problem, how do I solve it now using elisp?".
So basically, your develop the solution bottom up. You can pretty much write an elisp function anywhere in emacs (in the very file you are editing, in the already open scratch buffer, in an elisp popup shell) and test the function incrementally. Also, the extensive in-built help system lets you very quickly lookup functions that you need.
Once you have something that solves your current problem, you can save it, if you want. Over time your custom collection of functions improves and then you decide if something there needs to be generalized or made into a package.
Obviously this needs some basic elisp knowledge and then just practice over time.
I was happy with the git cli (and occasional tig) when I started using magit full time and initially never understood all the buzz. It did not seem that powerful over what I was using.
After 3-6 months of use, for some reason I had to go back to the cli and then I realized the ease and speed I had got accustomed to.
Magit is magical; use it for some time and try going back to your previous setup.
1. Software: this is all Pytorch/HF, so completely open-source. This is total parity between what corporates have and what the public has.
2. Model weights: Meta and a few other orgs release open models - as opposed to OpenAI's closed models. So, ok, we have something to work with.
3. Data: to actually do anything useful you need tons of data. This is beyond the reach of the ordinary man, setting aside the legality issues.
4. Hardware: GPUs, which are extremely expensive. Not just that, even if you have the top dollars, you have to go stand in a queue and wait for O(months), since mega-corporates have gotten there before you.
For Inference, you need 1,2 and 4. For training (or fine-tuning), you need all of these. With newer and larger models like the latest Llama, 4 is truly beyond the reach of ordinary entities.
This is NOTHING like open-source, where a random guy can edit/recompile/deploy software on a commodity computer. Wrt LLMs, Data/Hardware are in the equation, the playing field is complete stacked. This thread has a bunch of people discussing nuances of 1 and 2, but this bike-shedding only hides the basic point: Control of LLMs are for mega-corps, not for individuals.