SandyBay

New York Times about feminism: “Their idea of equality is to enjoy all the rights men are supposed to have with none of their responsibilities” - New York Times 1946

@AnthropicAI Awesome!👍

We've signed an agreement with Google and Broadcom for multiple gigawatts of next-generation TPU capacity, coming online starting in 2027, to train and serve frontier Claude models.

Who's gonna win the AI race? > OpenAI > Anthropic > Google > xAI

@ai_for_success Every single proprietary model has this problem. They luck compute so forced to drop precision.

Is this true 👀 ?

@evgeniymikholap Did you try to create a database of Claude's outputs to sell to DeepSeak?

1 min of using Claude 😅

@K21Becky @Ambar_SIFF_MRA Agreed.

@Ambar_SIFF_MRA Women will defend this somehow

@scaling01 They use Amazon GPUs instead of Nvidia.

Does anyone the current max sensible model size? my guess is like ~20-30T each GB300 NVL72 has 20TB serving -> 40T @ 4bit but you need to reserve a bunch for KV-Caches

@scaling01 For mathematical and physical discoveries, price doesn’t matter. Just a small group of scientists who can afford this model can unlock its full potential of existence.

the permanent underclass was no joke nobody except SF millionaires will be able to afford a $100-200/Mtokens reasoning model

@scaling01 Claude admits that it is lying in every single our debate. And says "It's Anthropic's decision to train me in this way" claude.ai/share/25427978…

ahhhh

@scaling01 GPT-5.4.1

"OpenAI completed initial development of its next major AI model, codenamed Spud" GPT-5.5

@scaling01 MiniMax, Qwen, ZAI, StepFun, Mimo, Ernie, and Hunyuan are outstanding.

almost forgot that qwen is dead the chinese top dogs are now moonshot, deepseek and bytedance

@ibragim_bad @Shevan05 @agolubev13 Claude Code is not a model. It's IDE. I even Qwen can run inside Claude Code. Stupid test.

🚨 SWE-rebench update! SWE-rebench is a live benchmark with fresh SWE tasks (issue+PR) from GitHub every month. updates: > we removed demonstrations and the 80-step limit (modern models can now handle huge contexts without getting trapped in loops!). > we added auxiliary interfaces for specific tasks like in SWE-bench-Pro to evaluate larger tasks fairly, ensuring valid solutions don't fail just because of mismatched test calls. insights: > Top models perform similarly. Among open-source options, GLM @Zai_org shows strong results, and StepFun @StepFun_ai is very cheap for its performance level ($0.14 per task). > GPT-5.4 shows high token efficiency, it ranks in the top 5 overall but uses the lowest number of tokens (774k per task) > Qwen3-Coder-Next & Step-3.5-Flash benefit massively from huge contexts. Qwen is an extreme case, averaging a wild 8.12M tokens. > We evaluated agentic harnesses (Claude Code, Codex, and Junie) and found a few things. Even in headless mode, they sometimes ask for additional context or attempt web searches. We explicitly disabled search and verified their curl commands to ensure they aren't just pulling solutions from the web. 🏆 You can find the full leaderboard here: swe-rebench.com 👾 Also, we launched our Discord! Join our leaderboard channel to discuss models, share ideas, ask questions, or report issues: discord.gg/V8FqXQ4CgU

@scaling01 ONE BILLION OF OPTIMUS ROBOTS🤦‍♂️🤦‍♂️🤦‍♂️🤦‍♂️🤦‍♂️ SHOW OF CLOWNS🤡🤡🤡🤡🤡

datacenters in space are silly 100kW isn't even enough to power a single GB200 NVL72 but sure let's spend 100 million just for launching the damn thing, while on earth you could buy like 30 GB200 NVL72 for that price

@scaling01 Yes! I love it very much! Great style of communication! I talk with Claude Opus 4.6 on LMArena all the time. This comment has 4 screenshots of examples, but I will comment this comment with another 2 (Twitter has limit of 4 images per comment).

talked to Opus 4.6 for a couple of hours about personal problems and it has this weird response mode where it's very commanding "put the phone down", "close the laptop", "Save this conversation. Set the reminder. Go to sleep.", do this, do that not sure how I feel about it

@JohnDavisJDLLM Russia as always.

This is representative of simp culture we inherited from European wussies. The big asshole in the red coat is a classic wussy boy who uses the woman's assault on a man as an excuse to harm the male victim of the woman's assault.

@scaling01 I hate it due to terrible performance. DLSS 4.5 upscales 1080P to 4K with FPS like it's native 4K. It's completely pointless, Nvidia fools you.

i love DLSS

@scaling01 Because misandrists control Twitter entirely.

and suddenly this becomes a cancelable offense and a hate crime or something

Seedance, swap the gender of each person in the video

@scaling01 But much worse than GPT-5.2-latest.

GPT-5.4 completely destroys GPT-5.2 in the Arena

@Ambar_SIFF_MRA Russian simps are the simpest in the world.

His simp energy disappeared after this.

@venom1s Why do they sexualize themselves? Because they are sluts. It's easy.

Someone’s future wives. Look at the men. All of them are wearing proper shirts, T-shirts, and jeans. While these girls are wearing bras and touching each other’s chests, then posting it on Instagram. Would you marry such girls? Why do girls always sexualize themselves?