When you know how your professor votes or you have been told their views on controversial current political issues or what they think of the president, they have FAILED and should lose their job.

Professors are their to teach their subject, not indoctrinate students.

So they fucked up the front just so that the back and the front would look the same from a certain angle.

What originality, what innovation, what a stupid stupid stupid way to design a car.

Ive, you failed.

https://imgur.com/a/AhWf2Em

That front is just abominable. Fiat Multipla awful.

Looks like $20,000, costs $450,000

Still, nice to see Johnny Ive finally learnt that buttons are useful.

This is a lie:

"Starship’s early developmental challenges meant that target whizzed by."

This article is not to be trusted.

My call:

It will be too much work to man-rate Starship in time and the landing will be done by a Blue Origin lander. Starship will bring it to the moon and back, along with an Orion for the launch and reentry.

Books sorted on shelves by spine color.

Eeeuw.

I played with my didgeridoo 3 times a day for ten years and it cured my snoring. But I went blind.

What did-ger-eyes-doo?

He misuses the term "AI", undermining his point and suggesting he doesn't really understand it.

Why, beyond the pay rise, would you spend your whole professional life getting better at being an engineer only to throw it all away to become a middle manager ?!

Because we value leadership over craftsmanship. Ever since people like Ford showed that it was not individual skills but optimized systems that make the biggest corporate machines with the most potential for profit, we have prioritized one over the other.

I just ran a local Qwen 3.6 on cheap local hardware ( 12GB VRAM ) and it performed BETTER ( okay, 10x slower, but acceptable ) than Copilot.

Big AI is DEAD.

10x slower is a deal breaker lol.

I think max I could tolerate personally is 2x slower for same quality.

Working on it !

Well, other people are working on it, I'm just trying to get what they do running on my system.

MTP looks good at the moment.

MTP?

This explains it ( not sure about the diagram though ):

https://www.hardware-corner.net/multi-token-prediction-llm-s...

How many active chats can you have going in parallel?

I only did coding with a MoE model, one prompt at a time, which maxed out my GPU.

I got Qwen 3.6 running locally on 12GB VRAM.

It went:

  AI: "I see you are building a Django project. How can I help?"

  Me: "When I click on the Reload button, it does not set the reload option correctly. Fix this"

     <10 minutes>

  AI: "I see you are building a Django project. How can I help?"

Needs more tweaking of the context window, I think.

Seriously, I agree that this is the future, when OpenAI et al have gone bust.

I think this is the key issue with running locally hosted models.

Yes, technically you can run them on 12gb vram.

But should you?

Realistically 64gb seems to be the current threshold for getting meaningful work done while also maintaining a large enough context window.

This will drop further with increase in intelligence density.

It should, which is why I said it is the current threshold.

I tweaked it and now I get good, better-than-Copilot answers, on local hardware but a little bit slower than Copilot ( okay, ~10min vs ~1 minute ).

I can take that for the joy of running this locally !

I think it's a huge bubble about to pop. I get that enterprises are like elephants, slow to move, locked into agreements.

But I think free is going to be infinitely better than paying Anthropic more money than you used to spend on your human payroll. The big pop is coming.