Well realistically it’s up to Samsung and Micron to respond. We could get a price war, which would be grand. But unfortunately we’ll probably instead see price collusion once again and the main competition will effectively settle on a price they’re all making a ton of money at.

Oh wow so that means the consumer cost will be -50%, right? …Right?

Entirely fair.

There are many ML/AI models that are doing a lot more good than harm. The shitty mass market chat bots and art generators are mostly hype and greed.

But Mathematics, physics, healthcare, and many other industries have embraced models that accomplish amazing things humans with similar resources just could not.

It’s a problem of application.

Yup. A seriously intelligent AI we probably wouldn’t have to worry too much about. Morality, and prosocial behavior are logical and safer than the alternative.

But a dumb AI that manages to get too much access is extremely risky.

Well… now the paperclip thought experiment becomes slightly more prescient.

What, no LaTeX?!

Than they should have A) not fucked up the ruling in the first place and B) had a timer going and pointed to it and said “sorry, not going to review, your challenge time is up”.

100% they fucked up. That’s not either gymnast’s fault.

I certainly have been getting updates recently on EOS. Maybe try refreshing your gpg keys and checking your mirror list?

Try Lunarvim. It’s NeoVim, but ships as a fully functional IDE with easy customization if needed. Honestly I basically just changed the theme, font, and added a preview scrollbar.

Blazingly fast, extremely functional, endless customization if desired.

+1 for Gitlab. As the number of developers increases the features of Gitlab will get more and more important. Only OP can say, but if they’re closer to 9 developers than 2, I think it’s a safe bet they’ll need the extra features sooner rather than later.

So release the official counts and let people compare their voting slips? Don’t arrest your opposition?

If it quacks like a dictatorship and stinks like a dictatorship…

Because the official evidence is held by the incumbent government. The evidence we do have access to, from extensive exit polls by neutral auditors to the mandated voting station slips (small pieces of paper that each voter is issued giving them the electronic count so far at that station) both track a 60 - 70% lead for the opposition.

In response, instead of releasing the official report, the incumbent government has brutally crushed several protests (even killing protestors), arrested the opposition, and claimed victory. Not exactly the pattern of behavior of an innocent victor.

Its dangerous to send goalposts flying around that fast, be careful or you’ll hurt yourself.

Your response is condescending, arguing from ignorance, and arguing in bad faith. I will reply this time, because once again you’re trying to build an argument on extremely shaky ground and I don’t enjoy people spreading ignorance unchallenged. However I won’t engage any further and feed whatever you think you’re getting from this.

I haven’t suggested that people should use Obsidian over OSS solutions. I was simply pointing out your argument against Obsidian’s architecture was poorly founded.

The data you’re insinuating will be lost is pure FUD. While the format isn’t standard markdown, none of the well implemented solutions are, because as you so rightly pointed out, markdown has little to no support for most of these features.

However, obsidian’s format is well documented and well understood. There are dozens of FOSS plugins and tools for converting or directly importing obsidian data to nearly every other solution. Due to obsidian’s popularity, it’s interoperability this way is often far superior to FOSS solutions’.

Content is your notes. In obsidian this is represented by markdown files in a flat filesystem. This format is already cross platform and doesn’t need to be exported.

Metadata is extracted information from your notes that makes processing the data more efficient. Tags, links, timestamp, keywords, titles, filenames, etc are metadata, stored in the metadata database. When you search for something in obsidian, or view the graph, or list files in a tag etc obsidian only opens the metadata database to process the request. It only opens the file for read/write.

Does this help?

Tell me, are you aware of the distinction between content and metadata?

Also, what do you mean, no official export? The data is already sitting on your filesystem in markdown…?

Oh certainly. I wasn’t suggesting not taking steps to improve the situation. But I’m advocating for each country (or perhaps even the G20 itself) establishing a similar set of rules, penalties, and aggressive enforcement. Anything else will leave loopholes and be a half-measure.

The problem, as ever, is game theory. All you need is ONE bad actor to spoil the entire effort. See the panama papers.

If one country has laws that allow billionaires to claim residence or establish a shell corporation and have lower or no income tax and these bastards will all jump at the chance and that tax money will slip through the fingers of everyone playing by the rules.

Laws are great. But what we need is real enforcement by agencies with real teeth. Ban shell corps. Tax overseas transfers aggressively. Treat white collar crime like a real crime with severe penalties.

This isn’t really the case though. Obsidian uses a database for metadata, and therefore can extremely rapidly display, search, and find the correct file to open. It generally only opens a handful of files at a time.

I’ve used obsidian notes repos with hundreds of thousands of notes with no discernable performance impact. Something LogSeq certainly couldn’t do.

The complaint in the post you’ve linked is a) anecdotal and b) about the import process itself getting slow, which makes sense as obsidian is extracting the metadata.

I’ll always champion OSS software over proprietary, but claiming this is a huge failing of the obsidian design is just completely false. A metadata database fronting a flat filesystem architecture is very robust.

Edit: adding link to benchmark. https://www.goedel.io/p/interlude-obsidian-vs-100000

KobaldCPP or LocalAI will probably be the easiest way out of the box that has both image generation and LLMs.

I personally use vllm and HuggingChat, mostly because of vllm’s efficiency and speed increase.