I was under the impression that when you're streaming the weights from disk because the full model won't fit in memory, that it is solely reading from the SSD, not writing, so it wouldn't be causing wear on your SSD.
The technology I was thinking of was digital screens. Analog clocks made sense when the best you could do was attach some arrows to some gears on a rotating shaft. Now we can individually toggle lights in a grid containing millions of lights.
> Digital. 1 on microwave less than 99. Where else in world 1 < 99 makes sense? Digital time math crazy.
> kids to learn how to read analog clocks [...] From what I understand that is becoming a lost skill.
I don't think kids need to learn to read analog clocks anymore, in the same way that we don't teach kids to use slide rules. Technology has advanced enough that analog clocks have joined polaroid cameras and vinyl records in the "obsolete technology that some people use for nostalgia or fashion" category.
(for the record, I grew up with analog clocks and I am fully fluent in using them.)
So many plots in Seinfeld would have been trivially solved with cellphones. Get separated in the parking garage? Call each other. Need help carrying an armoire? Call each other. Trying to meet up at the movie theater? Call each other.
I self-host my own code forge (a gitea instance) and I've mirrored a couple of large repositories on it. The traffic from AI crawlers was bringing my entire server to its knees. As soon as I put the large repos behind a login, my server's CPU usage dropped to ~5%. I don't doubt for a second that github is under a torrential downpour of AI crawler traffic. Those locusts destroy everything they touch.
You worded that as if you speak for Microsoft, but your account is 2 days old and this is your first comment, which makes me suspect you are a karma-farming bot.
I actually do, and I still notice it every time. Some asshole across the street rides a motorcycle and every day he revs his motorcycle which sets off the alarm of some other asshole's van.
> Another underappreciated feature about NixOS is, that you can create a VM out of a NixOS config to locally test a server update/deployment.
And similarly, you can generate a bootable ISO of your NixOS config. It's really nice for both testing things out and also as a recovery mechanism. Rather than booting some generic ISO, connecting to the internet, installing whatever recovery tools and filesystem drivers I need, and then doing the recovery, I can boot an ISO that is identical to my main system, with all my favorite recovery tools pre-installed.
One suggestion for improving the cache's posture: Give each dedicated builder machine a TPM. On each builder, generate a TPM-backed signing key. Only use those TPM-backed keys for signing packages that end up in the cache. That way we'd ensure that signed packages _must_ have been built on the dedicated build machines (due to the TPM) rather than potentially uploaded from a compromised developer machine using a compromised signing key.
It seems the current situation is there is a single signing key that has been in use since 2015 which implies that key is not tied to specific hardware and therefore could be leaked/exfiltrated/copied: https://github.com/NixOS/rfcs/pull/149/files#diff-ac4cc42774...
NYC taxis drive an average of 180 miles per shift[0] so they'd need to triple the range or have some sort of system where they can swap their depleted taxi for a freshly-charged one mid-shift.
I did some math, supposedly the complete install of the latest Call of Duty game is a 200GB download[0]. At 1gbps we're talking 26 minutes of downloading. At 2.5gbps we're talking 10 minutes of downloading. I'm honestly surprised game downloads have become so massive but are those 16 extra minutes really going to change anything?
Personally, I'm rarely "surprised" by a need to play a specific game that I don't already have downloaded/installed so I can just tell Steam to download the game in advance. But if I were to be in such a surprise scenario, we're talking the difference between popping on one youtube video while I wait or popping on two youtube videos while I wait. In both scenarios, I am waiting for a small but not insignificant amount of time... now if we could get 10gbps that'd be a game changer. I wouldn't even context switch for a 2.6 minute wait.