Salgat

3,152 karmajoined hace 11 años

www.salgat.net salgat.blogspot.net

comments

hace 11 horasdiscuss

You do get a public IPv6 IP, which is fine for most people (and with a simple script on a cron can keep a AAAA up to date, not that it changes often). And like someone else said, if you insist, you can use something like tailscale to punch a hole in Starlink's global NAT.

hace 11 horasdiscuss

Starlink is a problem that solves itself. If enough fiber rolls out that there's no more customers, they'll scale back satellites (since they only last 3-5 years).

anteayerdiscuss

Kind of refreshing though that the "throw more processing at it" scaling we saw in the 90s has returned in a different way. For a while we were really bottlenecked in our advances by relatively low levels of parallelism (most software used by your average user doesn't scale cleanly with more than a few threads).

hace 3 díasdiscuss

Anthropic wants companies that use API billing. Their growth period through hooking individual users who bring it to the business is over.

hace 3 díasdiscuss

Lots of small moving parts that build up into a large architected platform, which requires careful design, otherwise you get a big ole ball of "it mostly works when we first created it but don't touch it". And yes, bad developers are susceptible to creating the same thing. That's the whole issue with vibe coding, it creates something that works, but is brittle, because there are no strict architecture requirements around the entire design besides vague suggestions like "follow best practices", and as it makes mistakes, it just patches whatever it can to make it work, instead of fixing the root problem with the design.

hace 8 díasdiscuss

That's not what is happening. Claude isn't charging for the tokens it generates from the OCR on its side, but it's still processing the same number of tokens as if you had sent the text, just with the extra step of OCR on Claude's side. This is 100% a loophole that's burning extra resources.

hace 10 díasdiscuss

Using only Fable, if I do /usage it will show both Fable and Opus usage, so I don't think it's charging Fable prices for Opus code gen.

hace 16 díasdiscuss

This is a misleading statement. The "private data" is still largely publicly produced data that has been curated through private agreements instead of scraping, such as reddit posts/comments (this is the "third-party data agreements" that companies like OpenAI mention). And yes, there is still a lot of processing done on this data, which is the norm for preparing training data.

hace 17 díasdiscuss

Supposedly memristors would be ideal for this (and it would be reprogrammable), but then again, memristors seem to be the carbon nanotubes of the computing world.

hace 28 díasdiscuss

This is signaling to non-US companies that Anthropic cannot provide reliable access to their models.

hace 28 díasdiscuss

It also signals that Anthropic is a bad choice if you need stable access to their product outside the US.

hace 28 díasdiscuss

It's ironic isn't it? All the marketing of how dangerous and powerful Mythos is and the government went "bet".

el mes pasadodiscuss

The whole point of async/await (besides allowing synchronous style code instead of callbacks) is that it lets you yield the thread efficiently back to the executor (OS threads have a massive overhead in comparison). And if you're doing something that blocks the thread instead of yielding it, you can choose between either spawn_blocking or a thread pool.

el mes pasadodiscuss

You're still cutting your annual income in half though. That's pretty big no?

el mes pasadodiscuss

To this day I do not get why Intel doesn't just offer massive memory options for their cards. Just charge what it costs to add the extra memory, no upcharge, and they will never be able to keep up with demand. Cheap VRAM is enough to justify a lot of open source investment into challenging CUDA.

el mes pasadodiscuss

That was the main reason for the big hype around Memristors 15 years ago. High density, high speed persistent memory to completely remove the need for hdd/ssds, potentially even removing the need for external memory altogether. So frustrating that it still seems like we're a long ways from that becoming reality. There's some renewed interest in Memristors as they can simulate neural network connections in models, so maybe the funding will return for it.

el mes pasadodiscuss

We're planning to go to Valkey for exactly that reason.

el mes pasadodiscuss

Religion relies on assumptions that cannot be proven physically, secularism just avoids those assumptions (I say this as a Christian who finds secular biblical scholarship fascinating and helpful).

el mes pasadodiscuss

Even without encoding, as long as decoding is supported for AV2, streaming sites like Youtube can always transcode uploads. The encoder on mobile hardware is more of a nice bonus as long as we have an AV1 encoder available in the meantime.

el mes pasadodiscuss

My hope is that hardware improvements (better node densities every 2-3 years, better designs, etc) will pick up the majority of the savings for these companies in the future, assuming LLM performance starts to taper off with diminishing returns.