Kassuno

こんにちは！

[Hantao Yu @HantaoYu_Theory]

Turns out that your favorite Leetcode problems 3SUM and All-Pairs Shortest Paths both have faster (worst-case) algorithms (n^1.9992 and n^2.9995 time respectively). @LeetCode time to update your database 😂

Congrats to my cool advisor @firebat03 and @pinkfloydie! Also congrats to @claudeai.

https://t.co/XkrtFNju4d

[Kassuno @Kassuno1]

@AtharvaRNaik @maksym_andr Thank you for complimenting my argument. But who talked about how many papers a researcher has?

[Kassuno @Kassuno1]

@MingchenZhuge congrats! so op

[Deven Pietrzak @DevenPzak]

I think I just trained NanoGPT in 9.65 seconds!?

Yesterday, @hermanbrunborg used a connected longest exact match model, trained concurrently on CPU, to get the nanoGPT record from 39.9 seconds to 21.5 seconds, in an impressive open PR. By adapting their work, and generalizing the CPU model to a larger share of the tokens, I got the record down to 9.65 seconds, and submitted the PR this morning.

I built on three open PRs, in addition to my contributions: the main basis for my generalization was from @hermanbrunborg, and additional work from @nthngdy, @cyrusghane, and Daniel Monroe. Their engineering is extremely impressive.

Of course, all of our records are still subject to review. As I mentioned in my comment on Herman‘s PR (and in my own PR description), CPU usage is a gray area in nanoGPT. Their PR was the first to discover a clever way to effectively use CPU during the run, so I’m looking forward to the official reviews.

This was another very fun challenge! Particularly (potentially) breaking the 10 second barrier yesterday. So much engineering tact is yet to be explored.

Huge shoutout to the three authors of the PRs for their innovations.

@hermanbrunborg - I will also be at stanford for the weekend along with some of my cofounders, and sf for a couple days after, would love to chat more IRL, if you’d be open 🙂

[Kassuno @Kassuno1]

@e869120 congrats! orz
how good are llms now at theoretical computer science?

[Kassuno @Kassuno1]

@celestepoasts here's another one from 6 pro, it claims to be 97.82% accurate

[Kassuno @Kassuno1]

@HildeKuehne so basically how about you show one good, interesting, scientific paper in the domain of CV that can be beneficial to researchers

[Kassuno @Kassuno1]

@HildeKuehne hi, can you stop shitposting and share something useful for once? everyday you are hating on ai slop but your twitter is becoming merely slop hater which makes your content an even worse... slop.
I thought the role of professors here is to share insights for the community

[Hyperstition @hyperstition_cc]

It's official : We trained NanoGPT in 39.9 seconds.

By percentage reduction in training time, this is a greater improvement than the previous 45 world records combined.

The merged Github PR: https://t.co/CUsCrQL0He

Here's how we do it : https://t.co/HtNqWCXtB0

Credit @DevenPzak who did this alone while simultaneously playing baseball for the polish national squad, taking them to the European Finals.

[Kassuno @Kassuno1]

@HildeKuehne @iclr_conf very thoughtful of you. there are lot of issues in this world and definitely hating on a future postdoc who's trying to find the next step in his life is one of them.
can't believe how much toxicity AI has brought to academics

[Ilya Sutskever @ilyasut]

if you value intelligence above all other human qualities, you’re gonna have a bad time

[Manas Kala @Anonymanasensei]

Yesterday: Peer Review is dead.

Today: NeurIPS results out.

"7 papers accepted, my team crushed it."

Long Live Peer Review.

[Kassuno @Kassuno1]

@FakePsyho what other tests are you still looking for? previously it was competitive programming, maths, heuristics contests and finally astra being able to solve all pencil puzzles without any tools.
It would be great to know the improvements you are looking for

[OpenAI @OpenAI]

We’re sharing a solution to the Navier-Stokes Millennium Prize Problem, one of the deepest problems at the frontier of mathematics.

The proof was produced by a group of agents, using an OpenAI next-generation model significantly more capable than GPT-6 Astra.

The problem concerns whether the description of smooth three-dimensional fluid motion modeled by the Navier-Stokes equations can break down. It has remained unresolved for roughly 90 years.

[Jerry Tworek @MillionInt]

It is mindblowing to me how few people realize that their lives and everything they know will change drastically in the near future.

At this point, it should be pretty clear.

[Kassuno @Kassuno1]

@PI010101 I'm pretty sure nobody with enough intelligence would enjoy such thing 😭😭

[Kassuno @Kassuno1]

@sheriyuo 🫡

[Kassuno @Kassuno1]

@FakePsyho @FakePsyho

[Kassuno @Kassuno1]

@FakePsyho Today for example, I gave this to 5.6 Sol Extra High and it took 16 minutes to reply without using any tools... but it found a correct solution