freakynit

Solution: Don't become too dependent on small number of customers.

"Local inference can cause a large under-utilization of resources" ... did he mean the resources that users themselves have fully paid-for themselves?

Not to mention, I bet users will be open to "rent" their hardware for proper utilization, provided he can pay for it.... real question is, will he?

Is anyone noticing that from past few days, a certain level of "priming" has started with such articles appearing more and more?

Somehow most of the hackernews links related to epstein files end up getting flagged in under an hour, even when there are hardly any negative comments.

Lmao... https://latentstate.net/ubi-and-the-price-of-obedience/

These monitoring checks can themselves become signalling channels if triggering them produces observable side effects.

I freaking hate liquid-glass UI Apple has "forced" onto us. No freaking way to disable it. I hate it so much.

M1 Air guy here... I'm not looking for a replacement still. Almost 6 years now. Runs butter-smooth. Best tech purchase I ever made.

thank you...

Bro, I had to read this sentence in the context of the previous one, 3 fcking times.. and it sill does not make any sense to me.

I don't want to make this discussion politicized.. but man... Trump truly burned everything down that the US built over decades of hard work and careful planning. And these things are extremely difficult to build back, especially in a world like today's.

I have a little different take..

Top models, capable of discovering and inventing stuff, held back from public, and available only as for-rent research service (or whatever the relevant bureaucratic term is) to big corporations and government agencies like NASA, DARPA, NSF, NIST, DOE, etc. only.

I have a feeling that we, the public, will stop getting access to more intelligent models starting next year.. i mean, not the incremental ones, but, major ones.. like the one you got going from opus 5 to opus 5.5.

Damn!!

I hope this does not translate to suppression of any dissent or freedom of speech in general.

People heavily underestimate the number of tricks available to corporates for taking on "debt": https://latentstate.net/the-hidden-ledger/

Regulation should require model auditing, which should be fully automated, free(or extremely affordable), and time-bound, irrespective of who the client is.

What Altman, Dario and other such folks are trying to do is put a price on getting the model audited, by playing on these: price, timelines, red-tape.

Maybe a test-suite, kept continuously updated, similar to those benchmarks we already have for measuring the intelligence of these models. Like arc-agi, or terminal bench, or more.

eMudhra 1FA OTP Login Bypass: https://mastodon.catgirl.cloud/@49016/117475921584308861

Deepseek effect?

"If you want to claim inference is highly profitable in general, you're going to have to answer why it won't end up like supermarkets or airlines, both of which have bigger moats than inference."

- I don't... "inference is profitable" is as of now. No one knows future, neither have I commented anything on it.

The statement is valid as it stands.

That's right.

1. The link is for rented infra. That means the inference provider doesnt have to worry about deprecation.

2. Newer gen nvidia gpu's like GB300 are way more profitable.

70%+ net profits margins are for rented infra, that means no major upfont expense. Also, these margins are against 60% utilization only.

You can well imagine if you own your infra and can make utilization over 90%, where these go.

And then, my link above is for K3 model, which, compared to ds flash models, is highly unoptimized.

If I had a few million lying around, I would have invested in inference.

It's exceptionally optimized... 890 bytes per token. At standard 32K context, that's less than 30MB.

1. https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash

2. https://www.mindstudio.ai/blog/deepseek-v4-1-flash-specs-arc...

For the ones who are still wondering how so much profits, inference is a highly profitable domain: https://inferencex.semianalysis.com/profit-estimator

Mario one-shotted by deepseek-v4.1-flsh-1017 (in high mode, not even max) with sound and programmatically generated assets in under 5 minutes: https://mario-dsv4-1-flash-1017.pagey.site/

This is in "high" mode .. 40 seconds only: https://ibb.co/5XTRKNJv

Looks much better.

Kinda shows their team is more engineers than marketers.

I believe these are the last sets of super-high-end models available to public (open or closed).. these companies probably will now start to withhold top-of-the-line models for themselves, governments, and their super-high-paying partners only.

Those benchmarks can't possibly be real, right? Right? <insert padme 4-panel meme here>