Your curated collection of saved posts and media
“We want to build AI models that actually understand Japanese context” An interview with Sakana AI’s Tarin Clanuwat published by @Tokyo_Gov https://t.co/1Z3tu00yJ3 Read about our initiative to create AI that appreciates Japanese culture along with its rich historical contexts. https://t.co/B7QdHprGGh
ZAI is on fire. @Zai_org Hot on the heels of GLM-4.6V, the team has dropped two more models on @huggingface: - AutoGLM-Phone-9B: a 9-billion-parameter “foundation model for smartphones” that reads the screen and acts on your behalf. Small enough for on-device inference, big enough to feel like future. - GLM-ASR-Nano-2512: a pocket-sized (2B) speech-recognition powerhouse that beats Whisper v3 on English, Mandarin, and Cantonese - It even picks up your whisper. Grab them here: - AutoGLM-Phone-9B → https://t.co/0Bt0DpLhv7 - GLM-ASR-Nano-2512 → https://t.co/HMe7BpR9rT
STOP WHY ARE MY POSTS BLOWING UP Hhi Grace and Phighting fandom I’ll try to draw more frequently https://t.co/dbcL5z5umB
New paper from OpenAI: Training LLMs for Honesty via Confessions "In this work we propose a method for eliciting an honest expression of an LLM’s shortcomings via a self-reported confession." "we train GPT-5-Thinking to produce confessions, and we evaluate its honesty in out-of-distribution scenarios measuring hallucination, instruction following, scheming, and reward hacking. We find that when the model lies or omits shortcomings in its “main” answer, it often confesses to these behaviors honestly, and this confession honesty modestly improves with training. "
🤖✨ Agentic AI is NOT “just another chatbot.” It’s a new kind of AI that thinks, plans, and takes action on its own — like a digital teammate that actually gets things DONE for you. Imagine this: You say “Plan my Japan trip.” The AI: 1️⃣ Researches flights, hotels, activities 2️⃣ Compares prices, checks reviews 3️⃣ Books everything and sends your itinerary All you do… is pack your bags. 🎒🇯🇵 No micromanaging. No step-by-step instructions. Just results. ⸻ 🌟 What makes Agentic AI different? Unlike old AI that waits for prompts, Agentic AI is: 🧠 Autonomous — works independently without your constant supervision 🔍 Thinks ahead — anticipates problems before you even notice 📚 Learns & adapts — remembers what you like and improves over time ⚡ Takes real action — not just talks ABOUT tasks, it does the tasks This is not “question → answer” AI. This is “goal → complete solution” AI. ⸻ 💡 Why does it matter? Because soon, everything in your daily life can be handled by AI agents: 📅 Scheduling your week 📧 Managing your emails 🏠 Running your smart home 💰 Optimizing your finances 🛍️ Ordering supplies before you run out ⚕️ Tracking symptoms & booking doctor appointments It’s like going from doing life manually… to finally having a personal team that handles the boring parts. ⸻ 🔥 The big shift: Old AI = a TOOL you use. Agentic AI = a TEAMMATE that works WITH you. 2025 won’t be the year AI becomes smarter — it will be the year AI starts working FOR you. Welcome to the Agentic era. 🌍🤝🤖
@s_guilbeault It’s time for you to go back to where you came from! https://t.co/LFJSrQkCfn
— nomin 4 tweets au from Jeprinikol universe will be🔞🔞 https://t.co/yxG2vamNJS
TRANSLATION: #PorDee Layaaaaaaaagggggggg!!!!!❤️ https://t.co/HIlBJ9RFKL
1. Native Parallel Reasoner: Reasoning in Parallelism via Self-Distilled Reinforcement Learning 🔑 Keywords: NPR, Large Language Models, self-distilled training, Parallel-Aware Policy Optimization, genuine parallel reasoning 💡 Category: Knowledge Representation and Reasoning 🌟 Research Objective: - The paper introduces NPR, a teacher-free framework that enhances Large Language Models (LLMs) with native parallel reasoning capabilities. 🛠️ Research Methods: - NPR employs a self-distilled progressive training paradigm, Parallel-Aware Policy Optimization (PAPO) algorithm, and a robust NPR Engine to enable native parallel cognition without external supervision. 💬 Research Conclusions: - Trained on Qwen3-4B, NPR achieves performance gains of up to 24.5% and inference speedups up to 4.6x, setting a new standard for efficient and scalable agentic reasoning with 100% genuine parallel execution. 👉 Paper link: https://t.co/La4uXPrBrA
Native Contributions. #viral #podcast #viralvideo https://t.co/uX9jpDOMOi
We just turned in 305,968 signatures to give voters a say on politicians’ gerrymandered congressional map! https://t.co/EuIy0NZ2Wv
Pretty impressive achievement for a bunch of leftist nut jobs if you ask me https://t.co/EKsvKpsYtw
We just turned in 305,968 signatures to give voters a say on politicians’ gerrymandered congressional map! https://t.co/EuIy0NZ2Wv
I sure as hell need to wear native more for real 😌 https://t.co/S6DmVlO8Bq
#bcdtrade am accepting anything i don't have yet. same UN. just send a request. thanks! https://t.co/XtX3D6foy1
@Native2Crypto Gm Native. Good for the soul for you to be whole https://t.co/czrMnmJSVT
มาแว้วววว ทุกคนนนน ร้านแต่งน่ารักเว่อออ🥰 รอแจกกอวอยุนะค้าบบบ ใครว่างๆก็มากินหนมน้า น่ารักทั้ง2เซ็ตเลยง่ะ เลือกไม่ถูก😭 📍Cilic Tea https://t.co/S4LJV7ztF2
As far as only in SF things go, this is high on my list. https://t.co/69LNDfvGil
@wabi didn’t have an in app messaging system built. So this morning I mumbled some words at my phone and wMessage was born. Minimal messaging app with texts, links, pics, and soon voice notes. Best part, no read receipts. I hate them. https://t.co/IHSSwDnYoV
The only reason we know about this case is that the judge decided to lift reporting restrictions on naming the rapists, following applications by media organisations. The unfortunate truth is that this trend is almost certainly worse than reported, as sealed cases, victim intimidation, and lenient sentencing have skewed public perception into thinking these horrific incidents are mere aberrations. These state-sponsored PR campaigns to suppress the truth have done nothing but curb accountability and deepen public distrust.
ELON MUSK: “It needs to be sustainably multi-planetary, so not just visiting, but actually multi-planetary in the sense that if you have planetary redundancy, so if there were to be a catastrophe on one of the planets, the other planet would survive. Starship is capable of doing that for the first time in history, and no AI was used to create it.”
Why aren’t school children taught about slavery in the Ottoman empire? Is it because millions of their slaves were Europeans? Or is it because it was the British, the United States, and Europeans that were the ones that stopped it? https://t.co/gQwZiJeLre
American woman says “I’ve had enough” of the white hate pushed on the population “We're the only ones that are constantly having diversity shoved down our throats. We are a global minority being replaced in our own countries. And I've had enough” “If you would have asked me like nine months ago if I would ever be making pro white content, I would have looked at you like you were insane. Like, the idea of a white person being proud to be white is so foreign to most of us. And it's really only been conflated with evil and hate and maliciousness. And it's like you stop and you think about that. White people are really the only ones that are taught to hate ourselves, or the only ones that are blamed for every and all atrocities across the planet over all of time. We are the only ones that are taught to be colorblind. We're the only ones that are constantly having diversity shoved down our throats. We are a global minority being replaced in our own countries. And I've had enough.”
This is the book @elonmusk is referring to: https://t.co/GZkZUuELFY. https://t.co/NYGpK3YYWR
Why is such a false history of slavery taught in American schools?
White guilt is a PsyOp weapon designed to accelerate the extinction of Western civilization. It ends with us. https://t.co/vNsWotXpdu
🚨 LMFAO! Elon Musk gives the most ELON answer to "what would you do if you started over with $1,000?" "It's somewhat of an impossible dichotomy, because civilization would have had to have been destroyed or something, in which case $1,000 is not going to solve your problems." "And if civilization hasn't melted, then I could probably just tell people to give me money, which I've done before." 🤣🔥 @elonmusk "A lot of things would have to go wrong for that to be the case. It's like, am I just emerging from prison, perhaps, with a stipend? All my companies have been confiscated." "I mean, it would take armageddon, which hopefully that doesn't happen, like Ragnarok next level, and I lost." "I mean, it's impossible for someone to have this amount of knowledge, all the knowledge that I have, and then be dropped down to a low resource amount." "Because the reality is that either something truly catastrophic has happened, like civilization has melted, or I will be able to ask people to just give me money, and with the promise that I will have a high return, which is what I'm able to do right now. Like if you give me a dollar, you will get back much more than a dollar."
“The most important thing that’s going to happen in the world in the next 3-4 year is data centers in space” - @GavinSBaker We have the first Nvidia H100 operating in space at @Starcloud_Inc_ 🚀 https://t.co/cqQak4DwlH
👨👩👧👦 私のTwitter家族: 👫 親: @yagamiharuto @karaage_same 👰 配偶者: @kaedemisebaya 👶 子ども: @kyurupi1919 @ku9to1ut0en ➡️ https://t.co/xHhXiXMPhT https://t.co/RfouNf0OG7
昨日、今日とこちらにお邪魔しています。 皆さんとても素敵です🕊️ かむゐ Hiromi ☺︎ #姫が愛したダニ小僧 #剱伎衆かむゐ #KAMUI https://t.co/Hgsd3Mxjkm
【2026年2月18日(水)19時まで受注中！】 ▼王（キミ） イラストレーター：お久しぶり(@imlllsn) お久しぶり氏の描くオリジナルバニーガールが1/4スケールで登場 ご予約はネイティブストアへ https://t.co/ycFoDfv4x5 for overseas https://t.co/uEF2IDevNk https://t.co/uka2lcTHMy
@PandaTalk8 哈哈，赞了，我这个可以打广告吗？最好用的推特多语言回复插件，免费无广，和熊猫老板一样的傻作者出品。 https://t.co/LdBEAy6dtr https://t.co/U8aaslFvbT
NATIVEPATH - HYDRATE SAMPLE PACK For a limited time, try our Native Hydrate Variety Pack for FREE! AD https://t.co/EaCzj5Uzpd https://t.co/ZL1wzJJFOg
