supernovae

Training on business websites and marketing materials. There is such an overwhelming amount of that in the web that it shaped models trained on it.

It's happening on all sides.

Local LLM's are hit worse. Its about 6k for 5090 or 15k for an RTX 6000 and the Mac Ultra 256 is upwards of 12k.

Sure if you already have hardware, you can frankenbuild a system - but even the "affordable" dev stations of the DGX Sparks went from 3500 to 5k and upwards of 8k depending on vendor.

All the meanwhile, OpenAI pushed Luna 6 which is crazy cheap suggesting they have flash models to compete with Chinese models.

I just hope we see more open weights. Nvidia has NEMO but their license doesn't allow NEMO to be re-used on say, Apple or ROCm - Nvidia hardware only. Big ol MEH

mimo 2.6 is kinda dumb and over tuned though but not bad for a checkpoint testing their RL

I have to ask you the same question.

GPT is still 20 a month for most people, 50-100 or 200 for pros.

Meanwhile, for local LLM's - everyone is chasing hardware that is crazy expensiv eand to get around it they're leasing it or putting it on credit card. DGX Sparks are insane, and you need 2 for most people talking in this thread. Mac Ultra 5 is amazing but 6k minimum 12k for the build most want so many people lease it for 240 a month which doesn't even include electricity or time in setup so instead of talking about facts, we get into weird arguments like console wars where people give APple a 5 trillion dollar company 250 a month for 36 months and don't even own their hardware money "because they can run local models" vs just paying for output from their choice of frontier.

I say this fully loving local llms and embracing them, but the reality is, local llms have gotten so expensive and continue to get expensive while we keep talking about this "Threat" of apis - where there are a lot more than openai and anthropic available much cheaper and competitive priced.

Oh, and they don't work better than OpenAI or else we wouldn't even have these discussions.

Everyone keeps saying this, but I don't believe it's true. I think Luna is just an MLA or sparse architecture like all the flash variants and it's just cheap to run.

In any case, over the past few years, the only thing that has gotten more expensive is the hardware to run local models while API and Subs have gotten more affordable or feature rich while remaining same price.

They can compete because they have the compute to run the volume and if it's good on agentic work, people will be less incentivised to use other models for "Tasky-y" work.

It's literally increasing their market opportunity

It's not penalizing efficiency, it's taxing a commodity.

Everyone is efficient, thus, the efficiency has to support what we'd call our economy/society.

I can foresee a token tax in the future in some ways, but then that would kill private LLMs if hat came true so i'm not sure anyone has an answer, but neo-luddism seems short sighted.

I don't have the answer as it pertains to Uber, but the "marketplace" approach to things is pretty universal.

Ebay fees are upwards of 10-15%. Amazon fees can be as high as 45% Apple's app store fee can be 15 to 30%

Uber wants to get you coming and going by charging a Sub on the front end and splitting fees on the backend... it's like the worst of capitalism

It's too bad these models are built on the expectation of pirating the world

If you were running a shop, you would realize that nearly 100% of the fraud is "not chrome, edge, safari, or firefox"

It's unfortunate yes but that's what drives the threat signatures

I spent truckloads of money on home theater and audio systems and have been quite happy with my Sonos.

I have the Sub, The original big play bar and some 1s and I have a record player that plays to sonos and then upstairs I have an Arc.

I've probably got 10 years on this system now and couldn't be happier. I remember having to replace my receivers 2-3 times in a 10 year time frame from failed circuits to needing HDMI to needing a newer HDMI and all that jazz.

My play bar has just been optical out from the TV and it's supported 5.1 audio realy well.

I use the spotify play direct. I'm not a huge fan of the newer app and I wasn't impressed with them EOL'ing products.

With that said, it's been wife and kid approved. Everyone in the house knows how to use it and its been problem free. No wires for my cat to chew on.

There is also a diminished value of mobile phones today vs yesteryear.

* Twitter is dead for all intents and purposes * Lots of mobile apps are dead/dying and people aren't using them in the same way they used to out of privacy or concerns - such as checking in where you eat or tracking your exercise/fitness. (Nike really screwed the pooch here) and see twitter dead * New phones just push performance which isn't really needed when people merely surf the web or click a few apps while they take a crap

I'd say a lot of the mobile apps are dead/dying because they gave up on the app store paradigm and went subscription. Take myfitnesspal for example - where its 80 bucks a year to subscribe to use USER GENERATED CONTENT and it isn't shareable because its not an app purchase, but rather a sub. So many apps did this now that the family share is useless..

and lets not forget phones are getting expensive in plans rather than more affordable.. remember the days of cable? that's what carriers like verizon are doing where its hard to avoid disney, hulu or game subs or other crap when you just want a cell phone plan

only because your linux distro is doing it for you not because bash has git completion built in.

on windows copilot is built into the task bar with a popout window on right side bar showing output that can work with clipboard. You can talk to it, copy in an image or screenshot and ask for details too. if you prompt for PowerShell it gives you a code block with a copy button. (oh and also reference links to learn more... which is much nice)