It got me into thinking, since the genuine product also failed the AI test, what if someone made a fake product with a packaging that passed the AI test? It shouldn't be that difficult. All you need to do is to feed the photo of your fake product into the AI and keep iterating on the packaging.

Not suggesting a provider but if you're willing to get a Chinamod GPU, go get RTX 3080 20G or RTX 2080 Ti 22G. Get a couple of them and you can probably run Qwen3.8-27B at a reasonable speed.

I've got a RTX 3080 20G for $450 like a year ago. With llama.cpp, bf16 kv cache, kv cache offloaded to RAM, Qwen3.8 27B UD-Q4_K_XL, single RTX 3080 20G, I got 10 tok/s initially and it dropped to 5 tok/s at 50k context. I'm thinking of getting another Chinamod GPU.

Building a dual Chinamod GPU machine would only cost like $1000~$1500. It's not too bad compared with the alternatives!

Or, theres the option of a 32gb v100 (about $600USD on taobao etc), where you can get 1200 of prefill at 80 of decode: https://github.com/geoffwatts/ninfer-v100 - that's _really_ cheap inference, and it's not a modified card - you just need to add a blower or water block.

> single RTX 3080 20G, I got 10 tok/s initially

Add $200, buy a used 3060 and you won't need to offload cache to RAM. Yo'd have like 50-60 t/s with MTP enabled.

aww. I used to have a 3060 and I upgraded to this 3080 20G.

Maybe it's time for me to get a 3060 back again. :P

I run qwen3.8 on 5060ti 16G RAM. If you can do with Q3 and 64k of context. It works great at 25t/s

A year ago was prior to the memory cartel. You're probably off by 2x.

I've checked. I can still get a 3080 20G for $525, or 2080 ti 22G for $380. The price has increased but not too bad.

Probably the modding had prevented it from gaining too much price increase.

There's a problem with buffering tho. If the device's slow and unresponsive, and the user tapped the rotation button several times, it would be confusing if the rotation action happens 10 seconds after the user tapping the button. The user'd be left confused like "alright. So, has my input been taken?"

Now that I'm wondering. How does iphone mitigate this problem?

It shouldn't buffer them like the author describes. It should execute the button’s function immediately when pressed. This might mean to cancel the current animation and jump ahead, or it might mean to speed it up by the appropriate factor so it takes the same amount of time as it does for one button press. Either way is massively preferable to a button that swallows my input.

One way to deal with it could be “guaranteed interrupt” action (something like sigkill, just for UI action queue).

Other way could be to actually visually indicate action queue depth.

It's actually not difficult to setup at all. The game just work out of the box without any mod. However there's a learning curve. Find a tutorial somewhere online and follow it thru then you're golden.

It took me maybe an hour to learn the first time I played it. It's worth it! :D

Tho after finishing the game a few times it gets old.

That’s not what they said. The game does work without mods, but the economy is weirdly nonsensical (AFAIK you can push unnecessary amounts of cargo over long distances, and the game rewards you anyway) and I personally find it kind of pointless, compared to something like Transport Fever 2.

You're right. I misread. I guess I need coffee.

Yeah cargodist's indeed broken. You could milk plenty amount of money by just setting up two airports far away.

Maybe try Simutrans. It's got actual economy simulation. I tried it and I didn't like it tho. I couldn't make a profit in that game and kept losing. :(

Toki!

Simutrans requires a lot of care when initially setting up the network: get something small and profitable first, an industry chain that makes money. Then expand slowly, only when you have cash flow, connect everything for the passengers. If you find something is not profitable, sell the vehicles or send them do something else.

No. A coal plant will happily pay millions of dollars for coal hauled from the far side of the map, despite there being a coal mine right next door. Not like real life at all.

Somewhat related: in logging country, on several occasions I've seen logging trucks full of tree trunks pass each other in opposite directions.

Well, sorry a game mechanic from 1994 isn't that advanced.

Tangentially relevant. Someone made a snake game on URL bar: https://demian.ferrei.ro/snake

I got 35000, 18/13/9/9/6. Not my first language.

Interesting how literally everyone here's performing better than I do. Perhaps that's because I just clicked on the first option whenever I don't know about a word.

This. I don't see the point of constantly changing UI as an end-user. The old one just work. It works perfectly. Now that you changed it and thing breaks. :|

It's still broken to this day

Unpopular opinion but I actually enjoy using USB Micro B more than USB-C. That's because USB-C is much more complicated and there're non-standard-compliant cables floating in the market. USB-C also has some fancy mode like voltage selection. If it got screwed up somehow leading it to supply the wrong voltage, it could fry your Pico board.

The form factor. It's tiny compared with mini PC. Unlike other SBC, the software support's good that it just work as advertised (GPIO, graphic hardware acceleration, etc.). If you're using it for work, it also has stable supply chain / commitment that you can ensure to be able to obtain a unit in at least the next few years.

But you're right. It's pretty overpriced for its performance. If I'm after the GPIO I'd just use a microcontroller. If I'm setting up a headless server I'd just use pretty much any other SBCs or mini PC. I'm personally not a big fan of Raspberry Pi.

The cheap Raspberry Pi's still available as Raspberry Pi Zero tho, if that's what you're after.

Not sure on the performance but it might be possible to port this Mini Micro to those platforms.

Doubtful. Isn't Mini Micro build on Unity? That has much higher system requirements.