subarctic

I tried out deepseek v4 pro via a couple providers from openrouter, and it's always getting 429s. Are you running it on your own hardware?

I know very little about this but it seems like the kind of thing that can and eventually will be solved computationally, not by people figuring out what a parameter or group of parameters represent

I watched his speech until he got cut off by the guy saying he's out of order and I have to admit I really didn't follow what he was saying

Not sure I get your point, they're more fragile than manual toothbrushes not less. The metal connector part starts getting loose and it becomes loud and ineffective

lol

Did they have to write this publicity blogpost as part of the deal to get access to Mythos?

So, are they gonna add official support for this in the wifi standard someday? It sounds like the way that this works is a non-standard workaround

Maybe he or his lawyers thought he would fare better with a judge? Hard to say. I've heard that there are a lot fewer jury trial in Canada vs US but I'm not sure why.

I still don't have access to Opus 5. I'm on the latest version from Homebrew, I guess the update hasn't made it there yet

Is there somewhere they don't do this?

Ahh ok I just did a search on hn.algolia.com for fly.io and nothing came back with 10+ comments in the past year (until I posted this). But now I'm realizing they just don't include the url in the search, even though it says they do

Oh ok i thought they charged a flat fee to use the platform now but maybe that was a temporary thing. My account is a few years old so i wouldn't know, i still have a few apps running on the smallest vm size and it's always under $5

Thanks for clarifying, if your read is correct and they're just putting more control in the user's hand then this isn't an issue at all

Saw this on my RSS feed and was surprised to see it hasn't been discussed yet on HN - fly.io used to be on the front page pretty regularly with some pretty big comment threads, but I had to go back over a year to find anything from them with more than a handful of comments. I guess they've become less relevant on here since they stopped allowing new users to host apps for free?

Just saw this on my rss feed. Surprised to see this hasn't been discussed more on HN - maybe it needs a better title

Are they still doing that? I saw that months ago

Interesting. I guess it fits since Openrouter's main value is in consolidating payment to all these different model providers

Which they got from getty images

Maybe what they're actually worried about is liability, not safety?

I agree, I'm on ios now and that's not because it's "better", it's because of everyone else that has an iphone

Did they move away from github?

I find Philips Sonicare toothbrushes break after about a year, long before the battery gives up. Which i guess is another reason not having integrated batteries would be nice. But then they might not make as much money whenever I buy a new one so they probably wouldn't do that

I don't run one of these sites that has these issues so I'm really not aware of this problem. How can it be that sites are getting overwhelmed with scrapers that are just looking for training data? You only need to scrape it once to train a model, so shouldn't there be less traffic from this than there is from search engines?

On the other hand if the article is wrong and the traffic is coming from other ai uses (like an agent visiting pages on behalf of a user) then that would make sense.

I haven't seen it do it by itself that much so far. Even then, a lot of the time is spent waiting for some shell command to complete, which isn't inference. I guess being able to do parallel inference is like having multiple cores in a cpu, you can get by without it by just sharing that single core, and having it just gives you a performance boost under certain circumstances

> Most useful LLM work is done in parallel

I guess what I'm doing is not considered that useful then? I usually only have zero, one, or occasionally two things actively doing inference at a time, be it claude code sessions or one of the chatgpt/claude web interfaces, and i bet that's true for like 95% of people using llms. And anyway i bet even the hardcore people using a bunch of parallel agents would appreciate having access to local, private inference for some things.

You're obviously right though that cloud inference isn't going away anytime soon

You learn that you can build a version of Postgres that passes the tests and improves on these benchmarks in rust. You gain access to a codebase that does it which you can learn a lot from too. Definitely not zero like someone else said

Ya the second one should be Artificial Artificial AI. So close lol

What on earth, this is the first I've heard of this. I thought the EU was all about privacy?

I mean you can learn a lot. You can learn what's possible with less effort to build a proof of concept. It's kind of like you had another engineer do it for you, you don't completely learn how to do it yourself but you can still learn a lot with much less effort

I mean if it's actually just another Postgres, you don't have to worry about network effects as much because anyone could use it in place of Postgres. But sure, the more people you know using this, the safer it would feel to use it