actually useful and the ollama gui could probably even simplify this more.

That was exactly my same question. Then I finished reading the post. The reason is pretty clear, and written in the post: it is faster than ollama+mlx.

how much faster?

I was benchmarking different models, different engines, and different draft models, I posted a video on twitter, and people started asking about the setup in the final screen recording. So the blog post isn't so much "how a beginner should setup something" it's "here's the setup I posted in the video".

Original video: https://x.com/Freerunnering/status/2065275403548168398

And in the blog post there is a table showing the different speeds I got from different engines.

Slowest combo was 38.1 tk/s, and the fastest was 72.2 tk/s. All from "the same" model.

ollama is a wrapper on top of llama.cpp, and it makes llama.cpp slower, why use it?

Also Ollama has other issues (like forgetting what it really is - a wrapper).

... there is no mac micro, but there is an appletv

Locked down BS appliance.

with that new cheaporino macbook running on the same chip, i imagine it's only a matter of time before someone figures out how to install and boot it on tv hardware.

the root of the problem is that we have no data privacy laws.

buy a keyboard that support qmk or zmk firmware and remap it further than you might ever want.

I'm not sure I've ever met a good ceo...

I've done it on an index, higher res than the old vive and significantly less 'screen door'.

PPD (pixels per degree):

* vive ~10ppd

* index ~11pdp

* quest3 ~25ppd

* steam frame ~30ppd

* AVP ~34ppd

while the ppd between the vive and the index is similar, from personal experience the coding experience on the index is far more comfortable, perhaps because of the significantly reduced 'screen door effect'.

I haven't purchased a meta quest 3 simply because i have no desire to give zuck money or have any kind of meta account, but perhaps i will have to see if i can't find one on ebay or something for cheap.

I think apple needs to turn the page on it's 'not friendly for game developers' stance that they've had for a very long time. They've done some embracing on phones/ipads for sure but the m-series graphics capability is massive, they should be making investments with game engines to really squeeze performance out of their stack.

This might be a little difficult seeing as how the largest gaming engine in the industry for AAA (unreal) is owned by a dude that has sued them for their store practices but shrug.

the steam frame is apparently 440 grams.

The display part is 185g and headstrap adds 245g, which has headphones and battery at the back. Seems like it's well balanced, but might be too heavy. If it's comfortable it will be the first ever decent VR device. Assuming that they've implemented eye-tracking based UI like Vision Pro, and I don't have to shoot tiny targets to click, which is hilariously bad UI.

"Assuming that they've implemented eye-tracking based UI like Vision Pro, and I don't have to shoot tiny targets to click, which is hilariously bad UI."

Assuming that the Steam Frame isn't accompanied by a complete change to the current SteamVR experience that hasn't been so much as hinted at, alas, no, SteamVR is full of tiny targets to shoot. I've only ever used the Meta Quest 3S' native UI but the smallest targets there are generally significantly bigger than the smallest targets in the SteamVR UI. On the plus side, once you activate some of those small targets you can do some cursor navigation like a conventional UI, and having that option is a breath of fresh air... but it's completely inconsistent. You experience it as a bonus when it's available because it's not even consistent enough to "miss it when it is gone", let alone for it to be a consistent navigation method.

We may get the obvious eye-tracking upgrade but the targets are still pretty small, it's going to need to be very accurate.

Just switching the input method from shooting to eye-control would be fine for me, if it's accurate enough. Clicks can be made from the controller.

I found the index to be a bit on the heavy side but comfortable. I certainly put in significant time in several different 'flight simulators' (elite dangerous, star wars squadrons, etc) with it. No battery of course.

Ambient computing goes back far further than that... and we are getting there. I have digital control over most of the lighting in my apartment as well as several other items through various iot adapters (button pushers, stepper motors attached to my blinds, power outlet relays). Audio reactive agents like alexa/bedpodsiri/whateverrunsheygoogle can put you in control of all of that. Nothing is stopping anyone from building a audio reactive agent that dumps input into their AI model or models of their choice if they want it.

With the right set-top-box, you tvs become just another display for whatever you want to send their way. We are extremely close to startrek tng style spaces. I'm sure there is a github repo out there that may even be there.

Why do you think apple has the appletv, the apple watch, the phone itself is essentially a pocketwatch with an addressable display.

I think we have all the parts it's just now about knitting them all together and building up a ux that works for normal people.