As I play more with Agents like Hermes and Openclaw I've come to realise these truly are the new GUI.

I have Radarr and Sonarr running on my homeserver. I switched my model to cloud Claude, pasted the API docs of said apps and told it to make 'search, add, remove, update, and statusupdate' available in a small MCP.

It took 7 minutes, I switched back to my local Qwen3.6 model and I haven't touched the webinterface of Radarr and Sonarr in weeks. I just ask the model.

Everyone now gets a chat with my (telegram) AI bot in stead of relaying requests through me.

I have been looking into a decent local device. DGX Spark, Mac Studio etc... I think I am willing to spend on this, it really does feel like the 'iPhone moment' for me: I am not going back to individual front-ends for everything when my AI bot is a unified frontend for all API based software.

> in stead of relaying requests through me.

Overseerr is a thing.

How are you handling secrets? I want hermes to do stuff on the internet but I am not enthused about dumping the requried keys in .env.local or using process wrapper services like infisical yet. Encapsulating hermes in a docker sandbox feels slippery and I'm always left thinking if i've hardened my server enough.

I was hoping for an M5 mini and Studio ahead of time, but I guess I'll have to wait a little longer.

Maybe by the time they sort it out there will be an M5 Ultra Mac Studio with a full terabyte of RAM.

If I start the application process now I might have my second mortgage approved by the time said 1TB RAM M5 Ultra is available.

You own a house?!

The Neo as well, I just need a Mac to test with and finding one is ridiculous. Hackintoshes are hell to setup and run like crap. I tried https://www.macincloud.com/ and that was a waste of time. Someone take my money.

Ever considered second hand slightly older gens? Even M1 is still great for many use cases. E.g. often corps are selling them on Ebay, in pretty good condition.

Indeed, ton's of refurbished for $250-$300

Yea it does look like 5 year old M1's are going for $300 on eBay, but man that's painful a 5 year old machine for that much when you could get a new Neo or Mini for $600, if only you could buy them. I probably should just get the M1, test and sell it back on eBay. Thanks.

If you're going to do any kind of work on it I'd choose a 5-year old 16 GB memory M1 over a Neo every single time. 8 GB is what's painful. The CPU difference is very small anyway.

If you literally just need to borrow one, I'd just buy an Air from Apple directly and then return it within the 14 day window. I'll sometimes do this if I need an extended repair on my personal one, or there's a new mac I want to try.

This is unethical.

If the return policy explicitly allows "change of mind", I'd say it's in the gray area. Though ofc it isn't sustainable if everyone starts doing this. I assume there's a ((returns:buys)/payment identity) metric to ban the largest offenders.

Also, there should be some universally accepted way to have access to your data and a secure personal computer in the duration your device is getting repaired.

If a $300 Mac can do the job, a $600 Mac is overkill.

Neo isn't much better than M1.

Not sure what happened to you with macincloud, but I used scaleway hosted Mac mini M1 a couple years ago for a self hosted CI server, and it was working very nicely.

They don't really advertise that you can install nothing on the machine. It's super restricted.

I think you can at $50/month

I remember 4 years ago I was able to setup MacOS in a virtual machine. Maybe you can setup an Intel copy of MacOS on a qemu/kvm virtual machine?

I tried this, https://oneclick-macos-simple-kvm.notaperson535.is-a.dev/, it wasn't bad for 'one click install' and got the VM running, it was just unusably slow and burning up my machine. Seems like I need graphics acceleration in the VM which doesn't work on my Ryzen GPU - QEMU running on Windows, and I don't want to deal with dual booting. Literally would rather buy a Mac (if I could).

If you send me an email I might have a machine or MacOS VM on that machine you can use.

Thanks for the offer, think I'm gonna go the eBay route.

I don’t even think you can do hackintoshes anymore post Apple silicon right? I remember having one about 10 years ago and it was absolutely fantastic and ran really well. Wish we could do that now so I wouldn’t have to develop apps on my M3 MacBook Air, which constantly runs out of memory and is a huge pain.

You can for now, but in the very near future (macOS 27 I think?) Apple will completely drop support for Intel, and Hackintosh will be dead

I guess the sudden demand is due to OpenClaw? But most people will still use cloud LLMs, right? Anything particular with the Mac Mini that non-Mac lack?

Not just OpenClaw. The Mac mini is just stupidly good value for a desktop computer, and the RAM prices have only enhanced its appeal.

Apple doesn't make much of a fuss about it but their chip performance is laughably ahead of the other chipmakers.

The Mac Mini M4 gets a score of 3788 in Geekbench[0]. The top of the PC processor chart is 3395[1]. It's not even Apple's latest chip!

PC processors can only keep up by adding more cores, but real world performance in many workloads is enhanced by having a smaller number of higher performance cores.

[0]: https://browser.geekbench.com/mac-benchmarks

[1]: https://browser.geekbench.com/processor-benchmarks

If you remove the Mac filter, its performance is not even in the top ten

Which is obvious if you spent more then half a microsecond thinking about it, because apple silicone barely draws any power - it's performance is fantastic in it's niche, which is squarely within what a home user cares about - but it's not leading on benchmark performance, because that's not what apple designed it for

The reason its coincidentally good for local ai inference is also just down to the fact the embedded GPU has shared memory access to the system VRAM. That means low performance/throughput but large memory.

Which is great for home use, but once again not gonna top charts.

Which top 10 are you talking about? If you mean the top absolute geekbench scores, those are always with the assistance of cryogenic cooling.

Mac mini has first-class access to iCloud, photos, iMessage etc. So if you are deep in the Apple ecosystem you might prefer it for that reason. I have a windows gaming desktop that I could use as a server for openclaw/cowork but I realized I simply don’t trust that system enough to give it access to all the personal stuff I’m giving to the AI. I trust Anthropic and Apple. I don’t trust whatever junk is running on my gaming desktop.

If you want to run local models, another advantage is Apple’s unified memory architecture. The biggest Mac mini has 64gb ram and Mac Studio has up to 512gb. Compare this little box to what monster Nvidia gpu system you would have to buy to get the same memory there. And how much your PG&E bill would go up. That doesn’t account for the shortage of basic $600 Mac minis though.

An M4 mini is overkill just to run OpenClaw. I'm running it on a Pentium J5005 and it's running 20 other services in Docker. I think the main thing was many wanted it to be able to access iMessage. I think people dream of also using the mac to run the LLM but the 16gb ones don't have enough ram.

When they say 'due to openclaw' they refer to running AI models that openclaw uses, not to openclaw itself.

The shortage is for the 512, 256, and 128 models.

The basic 16GB Mac mini is also hard to buy. I bought one used not to save money but because I couldn’t find any store online with it in stock.

Those are the ones that can run the LLMs. Not a coincidence.

People are running openclown on microcontrollers.

You can look up benchmarks. It's different depending on the model of Mac Mini and Model of LLM.

The take away is that some of the Apple hardware hits a sweet spot for performance and price which may change in the future but for now it's causing a lot of demand so people can run inference without GPUs.

Also Macs keep a lot of their resale value so you can use them for a while and then sell them for sometimes 80% of their original value.

Affordable ram!

I recently bought one for my k3s cluster, and it was the cheapest 16g ram I could get by a decent margin.

My understanding is that openclaw is only a factor, and a relatively minor one.

Most likely the limiting factor is the crunch that chip companies are going through.

It is the SoC, not the memory as reported in the earning call. And lead time for SOC is 3-4 months. i.e even if they decided to increase order in March, if would be at least until July those volumes reaches warehouse ready to be shipped. And that is assuming there are spare capacity from TSMC for Apple to order, right now there is very little to none.

What annoys me most isn't the Mac Studio and Mini. It is the Neo. Someone must have done a poor job in demand planing. ( As well as pricing ). Only 5M unit till the end of the year when they are now increasing it to 10M. And it will likely miss this education's year cycle in the summer.

Hopefully they do better with A19 Pro Neo. Mac could reach up to 400M to 500M usage share. Roughly 25% of PC market.

Historically whenever they’ve done lower price products - iPhone SE, the current E editions, etc - they’ve sold poorly, that’s what probably got them.

The thing is that the Neo is actually useful.

I disagree.

I am old enough to remember the iPod nano -- Especially the 2nd generation. They were effectively low-priced and smaller iPods.

Apple sold millions of these much much quicker than the iPods and iPod minis (which came right before). Especially in 2006, it was _the_ "Christmas gift" just before the iPhone, iPod touch and later iPad mini took over. Possibly Steve Jobs' demo where he showed how they fit into the otherwise useless small jeans pocket helped convince the world.

The iPod nano effectively wiped out the competing music player market.

The Neo reminds me of the iPod nano and iPad mini. It is smaller and cheaper version of an existing successful product.

I think the iPhone SE and E are the outliers.

>It is the SoC, not the memory as reported in the earning call.

Is the memory not part of the SoC?

It’s a separate die, even in chips like the phone ones where it’s in the same package as the SoC.

Rightly deserved

More fomo

Opps, our lack of foresight into AI tripped us up, again.

This is just a preview.

At some point even the most economically liberal people will say that enough is enough. Making money and building capital is perfectly okay if you're working hard, but if you use said money or capital against the rest of us (who chose a different life) then we have a problem.

The Mac Studio definitely shouldn't be.

My M2 studio was the only computer I ever owned that had issues with the USBC ports not working with certain cables (and for the price, it should have had better performance).

I've owned a M2 Mac Studio, PowerMac G5, Mac Pro. Every single one had flaws that you would consider inexcusable on PC Hardware priced half that amount.

The PowerMac G5 had terrible video cards (the liquid cooled ones also had issues with leaks, but ignoring that). The Mac Pro also had terrible Video cards (they were PCI-X), but also Fully Buffered ECC ram (which cost substantially more than any other ram)..

Apple still can't even manufacture a proper mouse (who the hell puts a USB C port at the bottom).

It's ridiculous..

If Linux distro's had a way to integrate Android as a first class citizen (like IOS is in MacOS), it would greatly boost the number of apps available in ecosystem, and have a huge impact on MacOS I feel. Waydroid is good, but, it still is too clunky (I'd like to see something more like Wine for Android, where its native)

Why ... did you put so much energy into typing this out? That's a lot of energy to put into .... something you don't like and probably shouldn't care so much about.

Do you wish you could go back to macs?

But the Mac Studio can run LLMs reasonably faster better than non-enterprise-server setup. That is the only thing that matters at this point - it's an LLM accelerator, not a personal computer, at this point.

Valve's Lepton (Waydroid fork) might solve this when it gets released.

Low supply doesn't mean high demand. I don't think many people are buying Mac Studios, so they just lowered production.

> so bad indeed that they took a bet on the low price sector with the Neo and abandoned the powerhouses

Says this on a post about the powerhouses all selling like hot cakes, with many months long waiting times.

The Neo won't sell dozens of models they will take the low end laptop market by storm. I think your comment will age very poorly

This exactly. No other laptop comes close on price for the hardware you get. Yeah you may get more ram in a PC but promise you it won’t feel as fast when you’re using it day to day or have as good of a display or battery life.

In same price range you can get a PC that not only has more ram but also has better multicore performance, better disk speed and better port selection. Yes neo wins on build quality, trackpad, speaker, display and battery life but the PC would also allow you to install any linux distro.

The Neo is already considered a huge success and is the reason for the scarcity.

> so bad indeed that they took a bet on the low price sector with the Neo and abandoned the powerhouses

The Neo isn't just a bet on low prices - it's a machine that convinces people they can get away with less RAM. In the middle of a pricing crunch, why wouldn't you ship an 8GB machine like the Neo?

Its a win-win, Apple gets to ship a brand new SKU in volume despite the RAM crunch, and they get to punch into a previously untouched market.

I'm hoping that the success of the Neo and the RAM shortage makes people realise that 8GB should be enough for most tasks without constantly swapping.

That 32GB or even 64GB is considered a minimum to be able to run some word processing, chat app, fetch remote content, and display funny cat photos is preposterous. In terms of information storage, these are absolutely immense numbers.

The infinite treadmill of chasing for more RAM and then immediately proceeding to carelessly fill all of it at the first line of code is part of a deeper, wasteful, and self-imposed obsolescence process.

We don't need more RAM, we need more frugal software.