I have used Macs since the Classic era. My best Mac was a PowerBook G4 that could run Windows on a VM faster than most Windows machines at the time. My first MacBook was brilliant, but I have noticed a decline since then. My 6-year-old MacBook Pro really struggles nowadays, whereas I remember a time when people proudly said their 10-year-old Macs were still snappy even during rosetta.
Currently, Linux is the preferred choice for work. Windows 11 Enterprise is not bad when stripped of all social, news, and 360 ads overhead, but Microsoft is really trying hard to mess it up there too.
Edit: 6 years old, not 4, and also Intel Macbook, so due an upgrade for sure
Your M2 MacBook Pro really struggles? That is genuinely crazy, given that I use one as a daily driver and it feels just as fast as the day I bought it.
I think the Apple Silicon transition has increased Mac longevity far beyond the Intel or PowerPC eras, and I am quite baffled you think otherwise.
Is it an Intel MacBook? I think those sucked for obvious reasons outside Apple’s control (Intel getting stuck at 14nm for ages) which they’ve already fixed (by abandoning Intel).
The M1 MacBook Air was more powerful than the top Intel i9 MBP config if I recall correctly.
This brings back some memories. I often recorded live gigs from the radio. It was the best way to discover new bands and share them with friends at parties. The Internet Archive keeps on giving, what a great project.
Indeed, thanks for pointing this out and the links. With the excitement I misread that it was an MR from the fork to the main project.
I don’t think I’m able to fix the title though.
I find it quite exciting to read some results in an effort to understand if TurboQuant main ideas can be applied to model weights. There are other similar projects, so we’ll see, but it seems some of this fork results look promising.
TQ4_1S on model weights with minimal quality loss is really great. The MR discussion thread with results is specially interesting, with some models much more impacted than others in PPL increase, possibly size and model architecture play a part. Are there consolidated learnings from all the experiments? Thanks for this!
This is basically what antigravity (Google’s Windsurf) ships with. Having more options to add this functionality to Open code / Claude code for local models is really awesome. MIT license too!
All in all I think these projects are really great for communities that are unable to get online. There are some nice Linux education distros that would go together well.
Edit: 6 years old, not 4, and also Intel Macbook, so due an upgrade for sure
reply