seanmcdirmid

7 ms·

91.
by seanmcdirmid 2mo ago
But you should probably just get a suica in your apple or google wallet, you don’t actually need an IC card. I’m not sure if using a phone makes it less fast though, I couldn’t tell (we had to get a card for my kid since he doesn’t have a p
92.
by seanmcdirmid 2mo ago
How is that not just better since the power requirements are less? The whole point against conventional ACs and for what we call heat pumps is purely that they use less electricities (well, they can also heat and cool, so you don’t need two
93.
by seanmcdirmid 2mo ago
It’s both. You need the visa to enter the country and get the green card.
94.
by seanmcdirmid 2mo ago
Heat pumps are just ACs with a lot more energy efficiency. How does that make people hypocrites?
95.
by seanmcdirmid 2mo ago
No, what? Did they just do this to make Chinese models harder to acquire outside of China?
96.
by seanmcdirmid 2mo ago
> So just imagine, having a normal job in the US and you're asked to leave to renew your visa and this stuff happens. The only way I can see this happening is if you try to convert from OTP to an H1. I used to do some runs up to Van
97.
by seanmcdirmid 2mo ago
No, I don't care what you think about the devices, and I wouldn't have even bothered replying if it was just criticism. Its comments like "we need to make Apple go away" that piss me off, because you are trying to force
98.
by seanmcdirmid 2mo ago
No, and at $4000+ per 5090, I'm unlikely to have any experience anytime soon.
99.
by seanmcdirmid 2mo ago
I mean, I can tell you the same thing, to f*ck off. If you don't want to buy Apple products, don't buy Apple products! But don't you dare prevent me from buying them because your butthurt demands that Apple be destroyed becau
100.
by seanmcdirmid 2mo ago
I don't think the $20/month packages come with API access at all. I bought a refurbished M3 Max a couple of years ago just to try things out, I get 90 toks/s which is good enough for a lot of tasks. But ya, when DRAM prices c
101.
by seanmcdirmid 2mo ago
Where are you getting DeepSeek for a few bucks a month? I spend around $3/day on it, sometimes $4.
102.
by seanmcdirmid 2mo ago
This seems to be the favotitr solution for the anti-Apple crowd: make it so no one can buy Apple products. Misery loves company.
103.
by seanmcdirmid 2mo ago
Sharding a dense model using Tensor Parallelism (TP) across dual RTX 5090s has a significantly worse performance penalty over PCIe than sharding an MoE model. If you are using multiple GPUs, MoE is basically going to be your only workable c
104.
by seanmcdirmid 2mo ago
> "if your model can fit into the VRAM" can be true for the Mac as well. It is much more likely for your model to fit in large unified memory of a Mac than the smaller more limited memory of a GPU. Even going with two 5090s, yo
105.
by seanmcdirmid 2mo ago
I'm old and established (after spending most of my life not), but ya, I get what you are saying. It was a huge leap for me to spend $3k on a laptop. Before local LLMs became a thing, I had completely lost interest in buying anything bu
106.
by seanmcdirmid 2mo ago
I think the problem is evaluation. They can't tell who the good juniors are with respect to the new skills that they are looking for, so they go for experience instead. It doesn't seem sustainable though for obvious reasons. >
107.
by seanmcdirmid 2mo ago
I bought my refurbished M3 Max MBP with 64 GB for $3k a couple of years ago. Before Ethan I never spent more than $2k for a computer.
108.
by seanmcdirmid 2mo ago
10 grand for 256GB new Ultra sounds too cheap in today’s crazy DRAM market, it feels too good to be true.
109.
by seanmcdirmid 2mo ago
Weird, I’ve run LLM batch sessions for hours on my Max M3 MBP. It doesn’t get very loud, though I’m not getting anything close to 100 tok/s on a 27b model, I use a 35b MoE model just to get 90 tok/s. The fan comes on but thermally
110.
by seanmcdirmid 2mo ago
As far as I know, if you want MTP you need to serve it with either MTPLX or oMLX. oMLX is more stable and user friendly. I’m using Goose but have tested Codex and OpenClaw among others. Codex worked fine but I couldn’t get web search to wor
111.
by seanmcdirmid 2mo ago
Yep, somewhere between Chengdu and Siguniang (4 sisters) if I remember correctly (it could have been on a later leg on to daocheng, but I think it was really early in our trip). You get into Tibetan areas very quickly if you go west from Ch
112.
by seanmcdirmid 2mo ago
We really should go with a harsh Chinese solution. If you do drugs in China, you hope your dad is Jackie Chan or you are going to have a really hard life. And being caught homeless in a big city gets you a free escorted bus ticket back to t
113.
by seanmcdirmid 2mo ago
I can’t wait for cars to get native speech recognition and understanding models. Right now talking to a car is such a bad experience, but the local model tech is already at a point where we could already just have KITT in our cars without a
114.
by seanmcdirmid 2mo ago
The worst toilet I ever saw was in China, but at 5000 meters in Sichuan, it was two slippery two by fours that you had to straddle on over a big open pit (don't slip or lose balance!). That was such an awesome trip circa 2005.
115.
by seanmcdirmid 2mo ago
Weird. When I was living in Beijing, I made a mental map of all my bathroom choices when I wasn't at a luxury shopping mall, because they just weren't that convenient and available. This includes squats, which I included since sit
116.
by seanmcdirmid 2mo ago
Learning how to build emergent systems is also a skill kids should learn these days. The closest I got was coding up game of life for CSE 142 (intro programming). If stochastic gradient descent isn't taught in whatever CS Theory 101 is
117.
by seanmcdirmid 2mo ago
I get up to 90 tokens / second with Jundot/Qwen3.6-35B-A3B-oQ6-mtp, on a M3 Max with 64GB. MoE so it is not a dense model, but that means it runs faster (also, mtp helps). It is a 30GB model, but you should be able to load it, oth
118.
by seanmcdirmid 2mo ago
I've had some success with: here is this android bug ticket in a product that I don't know much about, fix it! And it works sometimes. I'm still technical, so it isn't the same, but it was crazy that AI could just solve
119.
by seanmcdirmid 2mo ago
Did they license an actual chip design from ARM or just the ISA?
120.
by seanmcdirmid 2mo ago
There are probably just as many women as men into LinkedIn.

More ›