HN user
twitter: @uneventual website: unlucy.com email: lucy at unlucy dot com
funny to think that there was a blip of people downloading software over the radio in the 80s, then the internet happened and it was all over hardwire, and now virtually all software is downloaded over the radio again
my friend pointed out that Q5_K_M quantization used for the open source models probably substantially reduces the quality of play. o1 mini's poor performance is puzzling, though.
They do some monkey business with sgx to make brute forcing harder but yeah it's worrisome.
https://blog.cryptographyengineering.com/2020/07/10/a-few-th...
It’s basically the exact same pricing as the US. $1 ≈ €1 today and $1200 * 1.25 for VAT equals $1500.
I don’t think so! Even permanent and supposedly non-arbitrary boundaries are subject to intense political contestation. Observe how in the first decades of this country’s history states were often admitted in pairs, one slave and one free (for instance Missouri and Maine), in order to maintain the balance of power.
Well to take the most obvious problem, each California congressional district has to have roughly 750k people but Los Angeles County has 9.9 million.
I continue to think that /the/ critical problem with web3 as pitched is that users cannot be trusted to maintain private keys in the long run. Either the keys need to be in HSMs and easy to rotate (à la Urbit's identity-as-NFT scheme) or you need a social/institutional escape hatch when the cold logic of cryptography fails you so you're not completely hosed. Maybe you need both.
It's a mixed picture, but to me the fact that irangov.ir still resolves in spite of an all-out siege by OFAC tells me that it's decentralized enough for almost anyone.
Oh that's awesome, I'm so glad people are hacking away at this :)
This is a solved problem, but nobody realizes it yet.
We already have a decentralized system of unique identities, and we have since the '80s. It's called the domain name system. They're human-readable, they have strong guarantees about being able to own and control them (so long as you pay a nominal fee), and they have a rock-solid infrastructure behind them that backs everything from Google to the US government to your friend's blog. We even have a (somewhat less convincing) way of verifying that the server you're talking to really is the one that your DNS record points to.
What still needs work is getting from an identity system, which tells you which server to look at for a given name, to a system of authentication for specific tasks. Given that someone controls a given domain name, how can they use that to log in to a service or post messages that are verifiably theirs?
If you're willing to run a server for it, OpenID works. If you only want to send email, DKIM has you covered. The w3c's decentralized identity specs are really cool, and I think did:web [1] has the potential to bring us to a world where you can buy a domain, cname it to some host, and upload your public keys there so that you can sign anything and login anywhere. Making this easy for non-technical users will be important, but I think it can be done. The fact that ICANN policy requires companies to allow you to migrate your domain guarantees that you can sign up with some fancy startup that will manage everything for you and keep your identity if you want to move somewhere else.