Hacker News Reader

> 8 gigs was too much of a compromise

i come from a ultrabook with 2 gigs of ram and the neo takes 2 vscode instances, one antigravity and firefox while keeping telegram and whatsapp on the background. that device can be used as a workhorse, pagination is very good.

can even run ios/android emulators but then you have to only have that project open, which is a non-issue for me. and the battery literally keeps working all day. good screen, good keyboard, good touchpad, operative system is close to linux. if youre thinking about getting one for programming and you're scared about the 8 gigs, hope i gave someone some light about it.

If the latest apple hardware could run Linux I would be buying a good amount. But I just hate the idea of being at the whims of macOS timelines. But I guess when you’re selling more than you can produce it doesn’t really matter.

> It's competing with a $200/mo subscription or renting server gpu time

Not if you're an enterprise that wants or requires on-prem inference.

if you're one of those enterprises it's perfectly reasonable to assume you're more likely to go have your procurement and legal teams negotiate with google for one of those weird boxes to run gemini on prem (https://cloud.google.com/blog/products/ai-machine-learning/r...) or other enterprise-y nonsense vs buying consumer hardware to run a chinese large language model in your network. I do not envy the person at a company having to get approval model by model because of weird open source licensing terms dealing with the "how do we know chinese models are safe" question (hopefully they're at least getting asked in the context of hooking it into an agent harness so there's some sort of plausible risk that necessitates the conversation - i can very much imagine it getting shut down to even run in a sandbox because chinese model + people being scared after the huggingface stuff). There's a lot of reasons to assume enterprise wouldn't be interested, and it's very very cool that they are IMO. Anthropic cut claude code rate limits - they dont seem to be see open source llms as a market threat (rhtetoric to the white house aside) yet but enterprises being willing to run consumer hardware to run local models could make things a bit more tangible

Wouldn't enterprises investing in their own hardware be more likely to buy standard 4-GPU or 8-GPU servers? Apple machines aren't really designed for server racks.

> another perfectly reasonable universe where 8 gigs was too much of a compromise and it flopped.

The target audience didn't evaluate that as a limitation - the majority of the market for Apple devices does trust that they will not produce and sell a computer incapable of support their use.

Professionals know there are tasks that a baseline computer cannot handle, and even common tasks that a more powerful computer does better, but those people weren't really the target demo for the laptop.

> The target audience didn't evaluate that as a limitation - the majority of the market for Apple devices does trust that they will not produce and sell a computer incapable of support their use.

Well, kinda the point. We know this to be fully true in retrospect (and many people correctly predicted it) but decent arguments existed against it at time of release.

>"the majority of the market for Apple devices does trust that they will not produce and sell a computer incapable of support their use"

the problem with that argument is that the vision pro exists, where they clearly overestimated demand, and where even among people with interest in VR and disposable income, the compromises on battery life and weight were actually too much to bear. Forecasting is just hard and you always need to be especially skeptical when you yourself are doing the forecasting for something you want to succeed. All those arguments for the neo line up in retrospect hindsight is always 2020

They also irritated plenty of devs, who aren't shelling out 3500 for what is in practical purposes a toy device, and then having Apple congratulate themselves the devs are the ones that should be happy they are allowed to play on Apple's kingdom in first place.

A lot of big enterprises struggle with ai adoption. Either they can't get models/tools approved fast enough, can't provision them in their internal network, or get slammed with exorbitant inference costs.

A big enterprise can drop one (or 4) of these on someone's desk and let them go nuts.

I think people just hate subscriptions so much they are willing to make an obviously worse financial choice to buy upfront.

I’d be willing to pay more to own vs subscribe, but the gap is currently far too large where buying a Mac Studio for AI is a straight up terrible investment.

I am also willing to overpay to not be reliant on cloud things, to have control of local hardware.

My biggest problem with buying a Mac Studio is that even in the case of the M5 Max models, I can’t think of any non-AI macOS applications in my creative life that have anywhere near that level of hardware requirements, and nor can I forecast that I would within its ordinary supported lifespan as a macOS machine. These machines left high-end stills and most video work behind generations ago, for example.

So while I would like a local machine that I can leave running in a way that I wouldn’t want to do with my old secondhand M1 Max laptop, it’s going to have to be something a little more pragmatic. Probably based around the Radeon R9700, since the AMD/Nvidia gap for LLMs is beginning to close, and even a single R9700 runs the main models I am interested in at speeds that are acceptably faster than what I have here.

I think the best move right now is to just wait a couple of years. There is room for the prices to drop dramatically. Buying the highest spec mac at the peak prices probably isn't a financially sensible move.

I tend to buy secondhand anyway, and yes, almost certainly it's best to wait.

At the moment, speccing out a local box more powerful than the machine I am using is largely an intellectual exercise, but it does have some value for understanding what a client might be able to use if they absolutely require on-premises inference, and I do have a couple who fall into that zone. So I am trying to keep up to date in principle, but the built machines don't get beyond the shopping cart.

The M5 Ultra has probably smoothed out the major issues of AI on Apple Silicon (prefill doesn't suck anymore) but the way Apple marketed that chip strongly suggests they are now fully aware of the strengths but also the limitations of their architecture and will get a lot more done.

It's properly put the (fairly equivalently priced) DGX Spark in the shade, though.