Dran

It means more software is about to go cloud-only. Decompiling Photoshop wouldn’t have any value if all the actual work is done via cloud GPUs and raw bitmap layers are returned instead. Double bonus for the shareholders you get software as a service and you also get to rent compute by the dollar on top.

I would unironically argue that a model primarily trained through distillation of closed frontier models, and then released open-weight with an open-source architecture, becomes “ethical” again.

Something something Robin Hood

I have nothing to contribute; I’m scared of used drives. Interested to hear who folks have had good luck with though.

You actually have a lot of options in this space. You can take the prefill engine of just about any LLM and turn it’s transformer into a classifier by lobotomizing out the decoder. You can also use a diffusion model on single pass to surprisingly competent result.

If Jeff looks easy enough to deploy by all means start there, but don’t discount the idea if Jeff sucks; you have a lot of options.

You could (and probably should) use a system-one style inference system for spam classification. Much cheaper and the structured output means it’s impossible to go rogue and curl some malware or whatever. It can absolutely misclassify but its output is programmatically structured and just ranks a pre-selected set of output tokens.

In your case that’s

Spam

Not_spam

What’s remarkable is that they can extend the trace length without DDR5 timings shitting the bed lol. Can’t even put 4 sticks in most boards without spooky interposers.

The unfortunate answer is because it’s cheaper to not give a shit. Sending a request and waiting for a timeout costs next to nothing, and scales linearly in terms of compute cost. The overwhelmed server on the other end slows exponentially with each concurrent request. The crawlers are set to maximize the efficiency of local resources, which include both wall-clock time and developer time. Why send one request at a time when your server can handle tens of thousands?

Try x; wait 60 seconds, if fail: put on a list to try again later.

Costs nothing to write and nothing to run. And if you own the hardware, and are paying for power already, may as well extract maximum dollar per watt.

+1 for pikvm. Expensive, but very powerful and very portable. My favorite part of them is their APL that lets me noninteractively, programmatically, interact with hardware buttons

Sittin’ outside of a Wendy’s for free wifi lookin’-ass

Win 11 IOT enterprise LTSC already exists and doesn’t even require a TPM.

Unless this is a legitimate channel for regular users to get IOT enterprise LTSC, this is a non-story. The thing we want already exists, they just refuse to sell it.

I don’t understand what you’re proposing that nagios/chrck_mk doesn’t already do

Copyparty is what I’ve been using. Seems very similar in philosophy.

+1 though for anti-discouragement. Competition breeds competence!

You can absolutely train a nontrivial task model on a gaming gpu (image classifier, sound classifier, text model with a very structured input and output, etc). You can also post-train addon layers on top of existing open-weight models (LORA).

I would argue that Linux kernel scale basically doesn’t work. It is a huge barrier to would-be developers to get into. You might argue that’s a feature, but it’s pretty hard to argue that it doesn’t add a lot of friction to the process

I’m glad I’m not the only one who is confused

Genuinely, pull requests being off platform without good integrated diff and merge conflict resolution tooling just doesn’t work at a certain scale.

I have no knowledge about the economics of live poultry, but rotisseries are generally loss leaders at grocery stores. They’re so commonly loss leaders that they’re often cited as one of the primary examples to explain the concept of a loss leader.

https://thehustle.co/the-economics-of-costco-rotisserie-chicken

https://en.wikipedia.org/wiki/Loss_leader

Is this hopium or did they release a statement suggesting an 3.8 35b a3b was in development for release?

I’m not aware of any public frontier LLM provider that uses a static seed for inference. Meaning, even with an identical prompt and identical model you will not get the same output. Seeds should absolutely come back with the streaming metadata on requests imho, but they don’t in any api/harness I’m aware of.

my guess is something like this

https://330k.github.io/misc_tools/unicode_steganography.html