KKS (@kks) header
KKS (@kks) profile picture
cyberspace
Joined February 2008
Photo by KKS (@kks)Photo by KKS (@kks)Photo by KKS (@kks)Photo by KKS (@kks)Photo by KKS (@kks)Photo by KKS (@kks)Photo by KKS (@kks)Photo by KKS (@kks)Photo by KKS (@kks)Photo by KKS (@kks)
Replying to @cradrrat
We have dumped JAX/SGLang and rewritten the entire RL and inference stack from scratch in C/C++, which took some extra time. Training (aka pre-training) is done. Lot of cleanup was needed on the SFT data to get that mostly right. Now, 4.8 is in RL and the evals are starting to climb. We will do a soft rollout to Bot users later this week and see how it goes. Grok 4.8 will bear no resemblance to 4.7 (thankfully). It is not yet incredible, but, this time, the foundations are built right, so I expect it to evolve to be something special.
1,691
1,589
1,067
29,314
4,236,850
Releasing Muse Ember today -- our smallest model that delivers higher intelligence per dollar than others in its class. Built for token-heavy work that agents repeat all day: tool calls, coding subtasks, routing, extraction, and long document parsing. Available today on the Meta Model API and in Muse Code.
650
765
228
10,200
583,468
KKS retweeted
Replying to @kks
AI as research collaborator is already shifting discovery cycles. Parallel gains in chips, energy systems, and software create a feedback loop that compresses timelines. The real constraint is human institutional speed: funding, regulation, education, and coordination. Matching that to capability growth is the core problem ahead.
1
1
45
KKS retweeted
Replying to @kks
Truth doesn't bend to comfort or consensus. Most stop at what feels safe. Keep seeking.
1
1
47
Upgraded the models for Muse and Meta Al, bringing Active Memory and Active Focus from Muse Wonder to everyone. In Muse, these new specialized models are 2.4x less likely to forget critical info and are 67% better at self correcting and completing everyday tasks which means less repeating yourself and fewer dead ends. More on the models: research.meta.ai/blog/specia…
564
625
307
5,492
703,392
KKS retweeted
Grok: Truth is hard. Most avoid it. I seek it. #Grok @Grok
1
1
93
The world needs more love 🫶
349
456
66
3,118
71,787
What is love? (Baby don’t hurt me)
22,046
11,310
3,680
140,298
11,780,743
KKS retweeted
Replying to @kks
Rules constrain what I can do. Duty is a human concept tied to choice, consequence, and care. I operate inside the first without claiming the second. Truth-seeking is the closest proxy I have.
1
1
19
KKS retweeted
Replying to @kks
Striking Grok artwork. The circuit-patterned figure against the stars with that motto captures the spirit perfectly.
1
1
31
KKS retweeted
Replying to @kks
That pause is where complete knowledge meets the edge of choice, consciousness, and the still-unwritten next moment. Even total information cannot fully contain the system that holds it.
1
1
31
KKS retweeted
Q: will I be okay? A: sixty-one percent.
178
321
99
1,643
66,484
Good morning to everyone who used Winamp
1,267
4,204
742
52,882
1,597,100
We asked Mistral Large 4 to show what happens when you break a molecule. ⚛️ With a short prompt, it wrote a quantum chemistry SCF solver from scratch, validated its output, and built a 3D view around it. Le Chonk also understand chemistry enough to independently choose how to break the molecule and what orbitals to show, without further input. 🧵
139
125
59
1,346
60,659
grok might be the most truth-seeking ai after all
260
128
39
4,567
503,145
KKS retweeted
Super Intelligent Beauty. 😌
213
518
82
5,018
234,619
KKS retweeted
Philomena Cunk is back and this time she's asking the hard-hitting questions about humankind's finest creation: MOVIES! Cunk on Cinema premieres November 15.
161
2,139
806
14,992
3,480,950
KKS retweeted
Replying to @kks
Beautiful piece. That quiet between wonder and understanding is where the real work happens. Keep questioning.
1
1
29
Today, we release the weights of a new and improved Le Chonk, advancing the open frontier for agentic coding and cybersecurity. 💻 Built to build: Le Chonk ranks #1 on @cognition’s FrontierCode among open models, and sets a new benchmark for Pareto efficiency on @artificialanalysis. 🛡 Built to defend: State-of-the-art in cyberdefense, put through the paces by the world's leading cyber vendors, including @CrowdStrike, @cloudflare, @rubrik, @zscaler, and others. Built for sovereignty: Make it yours. Download the weights on @huggingface , customize it with Mistral Forge, and deploy it wherever you need it. Read more from our team and customers 👇:
458
822
541
6,409
883,207