strnam

[CLS @ChengleiSi]

everyone should follow @yihaocs, he's a legend 😉

[David Feiock @davidfeiock]

In the 19th century, aluminum was more valuable than gold.

It was considered a luxury metal, prized for its rarity and appearance.

Then, in 1886, a breakthrough in electrolysis collapsed the cost of production.

Today, aluminum costs ~$1/lb and is everywhere.

Gold next.

[Yuandong Tian @tydsh]

Today, we release our first model, Flint-40B-A2B, a fast local coding model built with our automated research system. https://t.co/hL2yHS8pam

1. Strong performance. 54.9% on Terminal-Bench 2.1, 89.7% on SWE-bench Pro v2, and 74.2% on SWE-bench Verified.

2. It is trained from scratch (tokenizer->Pretrain->Midtrain->SFT->RL).

3. Fast in your local laptop. Up to 2x faster than comparable models: 125+ tokens/sec on a MacBook Pro, 150+ on RTX Spark.

4. We build the complete model training pipeline in 4 month!

[CLS @ChengleiSi]

I’ve learned so much about pretraining from @yingbozhou_ai, a real legend

[Quantumscapedia @Quantumscapedia]

Brilliant Q3 performance @QuantumScapeCo . 100MWH Solid State Battery production line coming online 2028. $QS to the moon 🚀

[Ratu Mataira @RatuD2]

We are on the path to a literal golden age.

Free electricity as a byproduct of our insatiable appetite for gold.

There will be a Nobel for this, and Adam will have to wait for his medal to cool down.

[Phillip Isola @phillip_isola]

We are in an era where you can agentify almost anything, and often you get a more powerful system.

This work provides a neat example of that approach applied to optimizers:

[Alexey Fateev @superalesha]

The guy who gave us QLoRA, says DeepSeek's inference system is on par with OpenAI's stack and better than what Anthropic runs, at least at their scale.

Now check the license. Apache 2.0.

Until today a production stack at that level existed inside a handful of companies. It was the thing that made their API fast and their margins fat, and none of them ever showed it to anyone.

Today any startup, any university, any cloud in any country can download it, read it and run it. It runs on NVIDIA Blackwell and on Huawei Ascend, so it isnt tied to 1 chip vendor either.

This is what I mean when I say open weights are about more than downloading a chatbot. DeepSeek is giving away the knowledge that only a few labs had, and that changes who can build an AI company at all.

Respect @deepseek_ai

[Dring Dring @tournonskiwi]

$QS Non, je ne re-rentrerai pas dans cette saucisse. Je ne re-rentrerai pas dans cette saucisse. Je ne re-rentrerai p… Eh, merde…

[Andrej Karpathy @karpathy]

@scaling01 Yeah I’d say the RL is working now all right

[Kevin Chen @Defiantclient2]

$QS: 🚨Mplus Moves Closer to Supplying Key Equipment to US Solid-State Battery Maker

Company developing assembly equipment and validating mass-production processes with @QuantumScapeCo

According to industry sources on Oct. 20, Mplus has developed assembly equipment for QuantumScape’s solid-state batteries and is conducting validation for mass production.

Beyond expanding its own manufacturing facilities, QuantumScape is pursuing a licensing business that would allow other companies to use its battery manufacturing technology. It has signed a technology licensing agreement with PowerCo covering battery production capacity of up to 85 GWh.

QuantumScape is also considering outsourcing production line operations to external companies. It is reportedly holding related discussions with South Korean battery manufacturers. If its manufacturing technology expands to external production facilities, Mplus could secure additional equipment orders, having developed machinery tailored to QuantumScape’s processes.

https://t.co/F81sMAQZer

[Antonio Lupetti @antoniolupetti]

"Mathematical Theory of the Transformer" is an extensive, 825-page mathematical monograph exploring the theoretical foundations of Transformer architectures.

It covers a remarkable range of topics, including embeddings and positional encoding, self-attention and multi-head attention, matrix decompositions, linear algebra, spectral analysis, information geometry, training dynamics, redundancy, optimization, and incremental Transformer architectures.

This is an advanced, mathematically rigorous read for anyone interested in understanding the mathematical structures underlying Transformer-based AI systems, including modern generative AI.

https://t.co/wwa1ucTYiF

[Mathelirium @mathelirium]

In General Relativity, a spinning Black Hole can lose energy when one fragment falls in while the other escapes with extra energy.

[Tri Dao @tri_dao]

We now have a platform for SFT & RL. This is why open weight models are great

[Ramez Medhat @R6mezMedhat]

The problem is that if there's no Ai they could have digged into some good blender or unreal engine courses and learned ton of stuff to make something like this
But they learned nothing and they have 0 courage to do so now

And that's a problem I personally face with AI in a daily basis,

The illusion of competence

[Tri Dao @tri_dao]

I’ve often wanted to pair a smart orchestrator / architect model with a super fast worker model. Diffusion LLM could fill the worker role

[Sasha Rush @srush_nlp]

Late to the party, but this blog post is excellent. Arguably the way people should first learn neural networks.

[Ruben @rdominguezibar]

An Anthropic researcher just told a room full of PhD students that their career plan no longer exists

His 3 pieces of advice for students:

1. Use AI every day. If you're on the exponential, you'll never fall behind it.
2. Choose projects that get better when compute scales. The rest will be solved before you finish them.
3. Stop planning your career as if it were stable. The safe option no longer buys safety; it no longer exists.

And a phrase that sums it all up:

"Writing is thinking. If you outsource your writing, you outsource your thinking."

The talk is titled "The End of the Anthropocene"

[Haiyu Wu @HaiyuWu1]

We are trying to build a free LeetCode for AI researchers. You can learn how to code core ideas from important papers across various areas (e.g., LLMs, JEPA, RL, diffusion), contribute to the problem bank, and importantly, post your thoughts to find like-minded people for collaboration or job opportunities.

Of course, the current version is still far from mature, so anyone who wants to help us with this project is welcome!

[Kent Langley @KentLangley]

I built a personal tutor for my son with the data set yesterday.

Because everything is linked in a graph it’s able to “learn” very quickly based on his interactions where he is overall and adjust itself.

I hope someone can extend the work to later grade levels!