Even if Ilya didn't really create this list I have a very good opinion about every paper on this page that I've read (most of them) so I think it's a great resource.
Lately during my off time I want to do something related to AI research (which I am already doing full time atm so I need something light) and I am for sure going to read through this.
Maybe I downplayed it too much but I really think this is still "in distribution" (we always have to remember that we are tech savy people and we influence the people that surround us). I see the value, but in my opinion it's not a generational opportunity, but a great acceleration. We are treating it like generational opportunity. That's why I say "everyone know there will be a crash, but noone knows how big that will be". The AI industry is not (in my opinion obviously) worth $ 391B [1] of added value.
While I agree on the smell I think that the situations are really different. I am not an economist but I think that other than the situation of the huge amount of money in play we are in a really different case. The general user (and I have noticed it especially with today's WWDC) basically doesn't get any benefit from AI (neither LLMs, image generation or photo editing). They were promised living like in Wall-e in 5 years and they are basically still living the same life. White collar jobs slightly benefited from the LLMs and same with programmers (while many say that they can get huge leverage the public results of what software companies produce didn't get the same benefit).
Everyone knows the market will crash, nobody knows how much.
In Italy they are really frowned upon by developers. They add 0 value. And it's not like "Oh, VC firms add 0 value to companies they acquire", this is really messed up.
Maybe they don't have enough racks. The news indicate that China isn't in a really good situation with GPUs, so probably they want to keep most of them for other stuff. Also because since the price is so cheap they probably want to use the other GPUs for stuff that has higher margins.
Yeah but as a European I think we took the wrong route. I am from Italy, and until 2001 we had 5 years undergraduate programs only. We then chose to do 3 + 2, but we should have gone with 4 + 1 years instead.
I have a BSc in Computer Engineering and I'm finishing a MSc in Computer Science. The MSc has been useless other than for being able to start doing research. I could have learned additional things in 1 more year, without repeating most of the knowledge in the other year, and then start the PhD directly.
Instead I did a MSc where for 1 year I mostly repeated old topics before starting working on really new things.
I think Masters should be highly specialized for people that after a Bachelor start to work but want additional knowledge for their position.
TLDR: 4 years Bachelors -> 4 years PhD is the correct route in my opinion. We messed up in Europe
In the comments many people are complaining that it's macos only, but one of the hosts say that they didn't just use screenshots to use the UI, but accessibility settings to see things hidden from what is visible on screen. So here's the reason... seems like that Apple's focus on accessibility will pay off and not help just people with disabilities.
as @goldenarm said just below here, I can't use Gemini Flash for how much it hallucinates. Gemini pro is quite speedy so I am good with using that. On another note, I really like deepseekv4 flash
I would suggest experts in interpretability (but everyone really) to go directly to the transformer circuits blog, where they explain their approach more in detail. Here is the link for this post: https://transformer-circuits.pub/2026/nla/index.html
Also, if you have never read it, I would suggest starting to read all the Transformer Circuits thread, by reading its "prologue" in distill pub