Tweet fixado
Exploring Direct Tensor Manipulation in Language Models: A Case Study in Binary-Level Model Enhancement: areu01or00.github.io/Tensor-Slayer.…
English
Tensor-Slayer
10.6K posts
@Eric26573545571 @jukan05 Will write , I have a few thoughts on this.
English
Yo Tensor. The CPU shortage is only accelerating; whats the play, still most bullish on AMD, or have news on Intel shifted your opinion? If agentic workloads are optimizied at 1:1 CPU to GPU ratio, perhaps the winner is the one who can manufacture the most CPUs (intel)? Not bullish ARM unless they increase margins?
English
>> 2025 Q4 Global CPU Market — AMD Sets All-Time Highs in Both Shipments and Revenue
- According to a report by market research firm Mercury Research, AMD achieved record highs in both shipment volume and revenue in the global CPU market in Q4 2025.
- Specifically: CPU revenue share reached 35.4%, up 6.8pp YoY and 2.9pp QoQ. CPU unit shipment share reached 29.2%, up 4.5pp YoY and 3.6pp QoQ.
- In short, AMD significantly expanded its share of the global CPU market in Q4 2025, with both shipments and revenue reaching their highest levels on record.
$AMD $INTC
English
I think both. Technological revolutions have historically benefit both type of workers. So I’m hopeful. I’m actually thinking the most prosperous time of humanity is here. Just need to leverage it right. The only difference is multiplier effect. One person moving how many needles at a given point. The high agency signal is clear.
English
I thought a bit about this. If AI makes humans a bottleneck, wouldn't the industries most impacted by AI demand exponentially more workers? That's basically what Jevons paradox would imply. Do you think the future is most bright for people in information heavy fields or physical jobs?
English
There is no other way to say this.
We have become the bottleneck.
English
People have a lot of great things to say about Hermes Agent.
We asked it to compile all this praise and make its own hype video.
It delivered:
English
@basil_k15_ Most probably not. But you can point your coding agent to the documentation to replicate the patch on latest version
English
@TensorSlay hey, one question, would this be able to receive the updates which openAI does to the official codex app?
English
Sorry for the spam. Situation has evolved quickly. Now, we have one click installer.
Steps to run to run Codex App on Linux :
1. Download the .dmg
2. Download one click installer and execute
Tensor-Slayer@TensorSlay
Codex App ported to Linux. Fully functional. How to reverse engineer and replicate detailed below :
English
4.6 is growing on me these days. It still has a different cadence and such but perhaps I judged it too harshly at first
English
English
Some updates: I've always been bullish on TML, and I actually joined TML this Monday
Looking back, I am feeling so lucky that I have the privilege to work closely with the best optimization experts on the Muon optimizer ( @Jianlin_S from Kimi and @clu_cheng from Meta). Now I am so excited to be able to work with @jxbz and build new cool things!
(On the other hand, there have always been some bad rumors about Meta TBD's potential failure. That's not true! From my personal experiences, it really has the best talents in the field, and I really enjoyed learning from the lab. The avocado model will for sure be great!)
JingyuanLiu@JingyuanLiu123
hmm I sort of disagree and I am bullish for TML. I think they really really have the top talents that I admire in the field, e.g. Jeremy and Sam for optimization, Songlin for Attn, Lia for MoE, Andrew for FSDPv2, and a bunch more folks it's just natural that it takes a while to publish good models: - dpsk starts to publish papers in 2023, even piblished dspkv2 (which I think is already amazing) in mid 2024 and nobody cares, until dpskv3 and r1 - msh took 10+ month to deliver a first not bad long ctx model in 2023 and be silent for the whole 2024 year, and starts to catch up gradually in 2025 - qwen starts to be a much better model than llama until qwen2.5, mid or late 2024, while the lab has been there forever it takes time to get infra and data done, but as long as you have good folks, and principled ways of doing science and experiments, some time or later, scaling laws will pay back
English
Those who don’t know who know
Kekius Maximus@Kekius_Sage
🚨 ANTHROPIC CEO WARNS: THE COMPANY IS NO LONGER SURE CLAUDE ISN’T CONSCIOUS.
English
@auto_grad_ A generational fumble
Xinyu Yang@Xinyu2ML
And the underlying reason seems to be that Alibaba has unified the entire team under the Qwen umbrella, with everything reporting directly to Alibaba’s CEO. That may have made the leadership above Junyang concerned that he could grow beyond their control, even though they rely heavily on Junyang and Qwen as core partners.
English
> Qwen 3.5 - Success
> Qwen 3 - Success
> Foundational frontier research - Success
> Multi-modal at scale - Success
> Small Language Models - Success
> Agentic Models - Success
> Reasoning Models - Success
> Ultra sparse MoEs - Success
>> mfw “asked to leave”
English
They have laid the foundational path and blueprint for other Chinese labs to accelerate upon.
Step-fun got you.
Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)@teortaxesTex
huh Well. Good job, babus *Alibaba* Qwen is pretty much over Ironic, it wasn't even a startup. Burned purely by lack of ethics
English
Thank you for your contribution to OSS. Qwen series has been phenomenal.
Junyang Lin@JustinLin610
me stepping down. bye my beloved qwen.
English
Words of wisdom.
<<Profile your crap. Optimize it. Test on ancient terminals.
It is not that difficult.>>
evilJazz@evilJazz
The next wave of enshitification is slowly rolling in with TUI apps eating more CPU cycles than actual GUI apps. Also wastes tons of cycles in the terminal displaying it. Profile your crap. Optimize it. Test on ancient terminals. It is not that difficult.
English
Going to have to get on this update
Pavel Durov@durov
All Telegram chatbots can now stream responses to users in real time — great for AI assistants.
English
0.8B - new model to experiment with. My earlier favourite was Qwen 3 - 0.6B. Great to validate ideas before scaling.
Lisan al Gaib@scaling01
Qwen3.5 2B and 0.8B Benchmarks almost seems like progress on tiny models has come to a stop
English
Time to replace ‘those who know those who don’t’ meme template
the tiny corp@__tinygrad__
TIL that Linux isn't free cause I had to buy the computer to run it 😭 This is the biggest load of cope I have ever heard. I hope Opus gets smoked by DeepSeek v4 and the only people who continue to use closed source models are Windows users.
English