smcleod

This resonates with me and I suspect a lot of others here. As always life is more nuanced that many would have argue.

Back in 2012 we used to give talks about how we ran all our CI builds in LXC containers, that came about because a bunch of us came from the BSD / Solaris world of jails and it just felt natural.

The folks at infracoders / DevOps meetups would think we were a bit "out there", then what do you know - 2013 rolls around and we quickly started to shift to Docker (too early some might say), but not long after shifting from LXC to Docker it seemed many others finally started to come round to containerisation.

Fast forward to 2026 and I've recently done some work with a client who is only just "getting into containers" (a Windows shop).

Thanks for your early work to bring better containerisation to Linux Solomon. It's funny to think we're 13 years from Dockers release. Time flies.

Both the AMD Ryzen 395/495 and the Spark have terribly slow bandwidth, their performance is slower than a M2 max, akin to just an M4 Pro or RTX 4060.

Why? Loads fine on iOS 27 Safari.

If you were only paying $20 for LLMs previously then you were getting very low usage / little done.

I wouldn't allow your terminal full disk access, that's quite a risk vector.

That's legitimately a good idea. I just contributed more to open street map in 5 minute than I have in 15 years.

I can't vote in America. The rest of the world is at the mercy of those within America that can as long as we're dependant on American products and services.

"The corpus, and the refresh that changed the answer" ... hello Opus 5.

I largely agree with this take. Unfortunately I think capitalism's political pressures have too much of a grip on society for there to be meaningful changes to the power dynamics that perpetuate the status quo.

Surely it's simply software that has its own kernel, that could be booted to from BIOS/uEFI or any other low level hardware initialisation process.

Yeah that can't be right, my M5 Max gets almost those speeds and certainly lot faster than what they're claiming the M3 ultra gets. Maybe they didn't have the model setup right or were running it with some unnecessarily high quant (>=8bit).

Considering how far backwards Google has gone since the release of the 3.x models, the brain drain they've let happen, the lacklustre software ecosystem and their track record with product management ... I would say they're more likely to drop trying to compete at the frontier and try to focus on something else instead.

How is that not a native thing with Actions? GitLab had multi runners since early on.

I work across multiple companies, but in general company profits are up across the board. Salaries are not. The larger enterprises are the ones that move the slowest and are the furthest behind. It's not just one thing (productivity), there's all of capitalism at play.

That's not really how it works. I'm able to get 5-10x more work done, perhaps more, it's a significant accelerant to those that know how to be productive. The problem is more for those who don't - and we bring them up to speed or help them pivot to other work.

The new releases and breakthroughs do the opposite for me - I feel energised by them. I felt like nothing truly that interesting had happened in tech for quite some time, now it's like the space race (except there is no one moon to reach).

I appreciate boring tech as much as the next well worn engineer and I'm not saying this is all positive but it's so sure as hell thrilling and you don't have to be an astronaut to immediately benefit (or suffer I guess) from it.

You're right to push back.

It all smells pretty load-bearing to me.

I wish someone would do this for Apple Music. Their app is dreadful!

Bandcamp, and I have on several occasions emailed the band directly and asked if I could purchase directly from them.

I'm running the UD-Q4_K_S on my 128GB M5 Max with 180K context, it uses around 100GB~

Not really compatible on all fronts, it's very capable especially with tool calling, workflows, logic and its base coding ability, but it's only a 27b model so it does not have anywhere near the level of knowledge baked in as larger models. This does not mean that it's not a good or useful model - it is on both accounts and very efficient, but it's not similar to a large model generally speaking.

No magic, just oMLX with MTP. You can look through the speed the community is getting here: https://omlx.ai/benchmarks/performance?model=qwen3.8&chip=&c...

Yes, I have the M5 Max. But there was no matmul acceleration before the M4 which made things a lot slower.

50-70tk/s is what I get on my m5 max on a 5-6bit Qwen 3.8 27B?

That was mainly before the M4 generation when they didn't have matmul instructions.

I wish you could drop $20k and get a house! Down payment here store like $100k+ (AUD). So "only" 5~ Max Studios.

It's very far behind llama.cpp, vLLM and SGLang in features yes. In part because of that but also due to some poor default settings it generally performs a lot worse as well.

I absolutely love TUIs, they can live in a pane in my terminal, run via SSH on remote machines, use hardly any resources and are very flexible.