smcleod

8 ms·

1.
by smcleod 7d ago
How is that not a native thing with Actions? GitLab had multi runners since early on.
2.
by smcleod 8d ago
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),
3.
by smcleod 8d ago
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
4.
by smcleod 13d ago
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 rea
5.
by smcleod 13d ago
You're right to push back.
6.
by smcleod 14d ago
It all smells pretty load-bearing to me.
7.
by smcleod 15d ago
I wish someone would do this for Apple Music. Their app is dreadful!
8.
by smcleod 15d ago
Bandcamp, and I have on several occasions emailed the band directly and asked if I could purchase directly from them.
9.
by smcleod 20d ago
I'm running the UD-Q4_K_S on my 128GB M5 Max with 180K context, it uses around 100GB~
10.
by smcleod 20d ago
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 larg
11.
by smcleod 22d ago
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...
12.
by smcleod 22d ago
Yes, I have the M5 Max. But there was no matmul acceleration before the M4 which made things a lot slower.
13.
by smcleod 22d ago
50-70tk/s is what I get on my m5 max on a 5-6bit Qwen 3.8 27B?
14.
by smcleod 22d ago
That was mainly before the M4 generation when they didn't have matmul instructions.
15.
by smcleod 22d ago
I wish you could drop $20k and get a house! Down payment here store like $100k+ (AUD). So "only" 5~ Max Studios.
16.
by smcleod 25d ago
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.
17.
by smcleod 27d ago
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.
18.
by smcleod 27d ago
That shouldn't be the case, it sounds like you've got something else going on with your setup. Here's my benchmarks: https://omlx.ai/my/fadc2127d384283f5df1fcc2c093a9f95700c6a52... which are inline with
19.
by smcleod 28d ago
AWQ 5bit, oQ5. oMLX.
20.
by smcleod 28d ago
It doesn't mean that, but yes it would (with a 5 bit quant).
21.
by smcleod 28d ago
I believe you mean 35B-A3B, there was no such thing as A4B. I use 27B and other models for software development, and quite a few research or similar agents. I cannot imagine a world where the old 35B-A3B model is smarter / more capable
22.
by smcleod 28d ago
I get around 70tk/s on the m5 max, with 5bit AWQ / oQ5 slowing only to around 40tk/s at higher context.
23.
by smcleod 28d ago
The smarter 27B is so fast with MTP I've found I really don't need the 35B-A3B. You get around 70tk/s on a M5 Max lowering to around 40tk/s at higher context sizes.
24.
by smcleod 28d ago
Unsloth use a property dataset they don't release, however you can indeed create quantisation locally on your machine and it's pretty easy, llama.cpp comes with everything you need.
25.
by smcleod 28d ago
GitHub's reliability has been going downhill longer than AI has been becoming popular. I feel like the load is becoming the scapegoat.
26.
by smcleod 29d ago
Musk is one of the last people I'd trust to centralise all my code with.
27.
by smcleod 1mo ago
I think Opus is just the new Sonnet, Fable is the new Opus. Introducing a new pricing tier is a killer way for them to raise prices.
28.
by smcleod 1mo ago
I think it's a bit inflated to label Gmail a "great tech for consumers". Maps certainly, YouTube mostly, Chrome & Android for loyalists, but Gmail hasn't been "great tech" for well over a decade. Additional
29.
by smcleod 1mo ago
Similarly there is research that shows the quality of LLM outputs strongly correlate with the education level (in the field) of the person promoting them.
30.
by smcleod 2mo ago
Japanese software and software technology culture is amongst the worst in the world. They're famous for failing to adopt modern ways of working, tools and as an industry is seriously struggling. There's fantastic video in this fro

More ›