how do they do "better by the day" anon? Are they literally deploying intermediate RL checkpoints?
how do they do "better by the day" anon?Are they literally deploying intermediate RL checkpoints?Hayden: @chetaslua give it another try
how do they do "better by the day" anon?Are they literally deploying intermediate RL checkpoints?Hayden: @chetaslua give it another try
a 15 yo Chiang today would be a little pink and say that Xi is a fat sellout cuck for not flattening Japan (and more than…
RT AestheticaSomeone tipped me off to a website where it compiles all the real ratings for movies on Rotten Tomatoes and compares it to the fake…
comments under a video about visiting a village in rural ChinaAll that overbuilt infrastructure has some perks, I guess(it's a «subway» for some reason. tbh looks…
"A mediocre researcher with a frontier model today is absolutely more capable than a very good researcher from a few years ago.""I am a 'mediocre' researcher…
I didn't anticipate that China has so much concentrated training grade compute at this point. A bit shook. Seems implausible in the extreme now that Dario…
RT MiraRe https://neonramen.kimi.pageIt's supposed to be good at frontend, so I had it make it a website on whatever it wanted.It chose a landing page for…
Rural revitalization W… Xi always wins in the endK: @voidsrus4 Chinese infrastructure is a meme because of how it's so bad that people don't live in…
RT Chamath PalihapitiyaTricking the US Government to protect frontier labs’ business model by using a China boogeyman is a mistake. It is protecting the equity of…
> these ghetto DROnES iNCoMiNG from red teamDoes Palmer actually prepare for war in Ukraine'2025Well I guess that's more than can be said of virtually anyone…
OMG China has discovered a cheat to build more gigawatts of compute faster (surprised)Zephyr: @teortaxesTex @ruima @tphuang DT is far more power hungry
RT Jitendra MALIKHighly performant open weights frontier models such as Kimi are a competitive threat to OpenAI & Anthropic, but probably for everyone else these are…
RT Wu Haoninga nice start & we still have much room to improveTeortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞): even K3 can ace IMOIt's using a lot…
It’s been this way for yearsWall Street Apes: My jaw literally dropped checking this outIf you go to New Jersey inmates mugshots and filter by race…
Grok is a solid workhorseMia: Still haven’t found a real use case where Grok 4.5 fails at something I ask vs GPT-5.6 Sol, Fable 5, or…
SpaceX’s massive corpus of world-class engineering data (excluding material blocked by ITAR) will be added during supplemental training of the 2T run.This will dramatically improve Grok’s…
I believe this is all broken telephone reporting99% sure there are not 524288 Ascend 950DT's in the world as of July 2026; certainly not at ZAI's…
RT Wu Haoningso bad boss xinyu i have never been or even seen gpu-rich🥹Xinyu Yang: True. When I came back from the US and joined Kimi,…
unironically this is happening right tf nowHamel Husain: http://x.com/i/article/2078343837609861120
And now from the Chinese government side.This would be a good time for cooperation between the US and China to establish common testing/acceptance standards for new…
holy lmaoI hope the meme about AI communism sticks. it's even better than the one with water. Zoomers are lazy and communists, they will definitely prefer…
how are bong journalists real? Kimi is now responsible for the selloff too? This is dumber than the original DeepSeek momentCarl Zha: Hell Yeah! About time!
2800/896*16 = 50 ✅1600/384*7= 29,166 ❌Bloomberg should learn arithmetic. Activated parameters cannot be calculated from expert sparsity ratio alone, at a minimum there's the attention backbone.…
What's going on here @zephyr_z9Bleys Goodson: Amazingly, when I crunched what it takes to serve Kimi K3, I found not only Huawei's next-gen Ascend 950DT on…