Are you interested in working on AI-driven discovery? Apply to our Software Engineer (Research and Development) position! https://t.co/E1uPJa5tIy

.@ResembleAI just dropped something HUGE today... and it might be the most advanced deepfake security tech I’ve seen this year. Meet DETECT-3B Omni. A single model that can detect fake voices, images, and videos, all in one unified system. Let's dive in 🧵↓ https://t.co/JkuvPNPP3r

Yep, I can bring your wildest dreams to life 😏 https://t.co/agTsZ8xtev

Yep, I can bring your wildest dreams to life 😏 https://t.co/agTsZ8xtev

+2 more

BREAKING: Artificial Intelligence is a heavy favorite to win TIME's Person of the Year Who should win? https://t.co/h5Q4qmaf2c

RT : I used to wake up to 12 tabs, 3 dashboards, and a Twitter thread for every hunch. I started routing my morning scans through @EdgenTech and my noise-to-signal ratio actually improved Their EDGM multi-agent model blends on-chain activity, social sentiment, and real-time analytics into one stream that reads like an analyst’s notebook institutional-grade insight without the subscription gate. Backed by Framework Ventures and North Island Ventures, it feels legit It flagged a clean trend well before the wider feeds lit up not hype, just faster context and crisper timing. Anyone else using @EdgenTech for their morning edge?

Now tracking @native_fi Credit Pools on our Yield Dashboard https://t.co/eaaMbrdH6B

The charts from the @ncri_io report shows some interesting anomalies. - The @NickJFuentes account receives more “Re-Tweets”, immediately after posting, than accounts with many times more followers. - With a majority of those being from “repeat early retweeters”. - Which are predominantly anonymous accounts which server no purpose other than to “retweet one account”. - Roughly half of which are located in non-Western nations. Considering all of the data and patterns… it seems highly likely that we are seeing a large-scale bot farm operation in action. How will @elonmusk & @X handle situations like this? Will they remove the suspected bots? Will they ban the account “benefiting” from the bots? If nothing is done to de-bot-ify X… then X will increasingly become a platform which does not represent reality.

+2 more

The productivity gap between male and female academics has increased after ChatGPT and it may be caused by men using LLMs more https://t.co/7vGpljLu2h

私がコミュ強陽キャだからではなく(それも多少あるかな)、サカナAIのカードを首から掛けていると色んな人から声をかけられるので自然と人と繋がれる。予想していたより広くグローバルで名前が知られていた https://t.co/p2R3bUmR50

私がコミュ強陽キャだからではなく(それも多少あるかな)、サカナAIのカードを首から掛けていると色んな人から声をかけられるので自然と人と繋がれる。予想していたより広くグローバルで名前が知られていた https://t.co/p2R3bUmR50

AI Scientist was featured at AlphaXiv’s event 🐠Looks like AI for Science is trending now #NeurIPS2025 https://t.co/mrFouRTEYP

@FrancisPSantora @hf0 This company: https://t.co/IjLFJIwgIN

i won the @xai hackathon by making an X feed for grok code introducing manaflowX. a curated algorithmic feed that lets AIs escalate the most urgent tasks for human review. built with @austinywang https://t.co/p9YdddHsTB

Apple chips chief Johny Srouji to his division this morning: “I love my team, and I love my job at Apple, and I don’t plan on leaving anytime soon.” https://t.co/qhJNq1Ny2g

Introducing Cluso, a unique take on local IDE for vibe coding. Conceived over 9 months ago, four false starts and demoed countless times, but finally it's here. Use @claudeai, @AmpCode, @kilocode. @cline, @cursor_ai, @opencode SDKs, APIs, Local, your own keys etc. FREE and Pro Edition (pro features to TBD) Beta this week, sign-up for early access and subscribe to my X and jump the queue; get a 15m demo with me https://t.co/ZKIHEO7TGe

Recently @OpenRouterAI conducted this research that shows the current 'State of AI'. https://t.co/7wEPv74pQT CEO of @Fetch_ai and chairman of @ASI_Alliance @HMsheikh4 comments: "The latest OpenRouter analysis offers one of the clearest empirical signals yet: LLMs are no longer merely tools for text generation, they are becoming behavioral systems with evolving patterns of use, agency and persistence. The shift that began with o1’s reasoning capabilities marks a deeper transition: we are moving from “models that answer” to “models that think, evaluate, and act.” Open-weight adoption, the surge of creative and coding use cases, and the emergence of persistent foundational cohorts all point toward a world where intelligence is not siloed behind proprietary APIs but is participatory, composable and increasingly agentic. At https://t.co/tLtcUosIij, this trajectory aligns with what we’ve been building toward for years. Our ASI:One model @asi1_official , designed for agentic inference from day one, is part of a broader belief that intelligence should not live in a vacuum. It should do things: coordinate services, negotiate, transact, and operate autonomously across digital and physical environments. This is the frontier we are entering. LLMs that evolve into agents, and agents that operate as an intelligent, decentralized mesh across the economy. The next phase of AI will not be dominated by a single model or a single platform. It will be an ecosystem, open, interoperable, and built around agents that can leverage models like ASI:One to reason, collaborate, and deliver real economic value."

Recently @OpenRouterAI conducted this research that shows the current 'State of AI'. https://t.co/7wEPv74pQT CEO of @Fetch_ai and chairman of @ASI_Alliance @HMsheikh4 comments: "The latest OpenRouter analysis offers one of the clearest empirical signals yet: LLMs are no longer merely tools for text generation, they are becoming behavioral systems with evolving patterns of use, agency and persistence. The shift that began with o1’s reasoning capabilities marks a deeper transition: we are moving from “models that answer” to “models that think, evaluate, and act.” Open-weight adoption, the surge of creative and coding use cases, and the emergence of persistent foundational cohorts all point toward a world where intelligence is not siloed behind proprietary APIs but is participatory, composable and increasingly agentic. At https://t.co/tLtcUosIij, this trajectory aligns with what we’ve been building toward for years. Our ASI:One model @asi1_official , designed for agentic inference from day one, is part of a broader belief that intelligence should not live in a vacuum. It should do things: coordinate services, negotiate, transact, and operate autonomously across digital and physical environments. This is the frontier we are entering. LLMs that evolve into agents, and agents that operate as an intelligent, decentralized mesh across the economy. The next phase of AI will not be dominated by a single model or a single platform. It will be an ecosystem, open, interoperable, and built around agents that can leverage models like ASI:One to reason, collaborate, and deliver real economic value."

The thing that I was most shocked by is she has a **really good** radio voice for narration. I don't know where she got these skills from LOL. Each episode is preceded by a short narration from her - to preserve the human element. So its not AI slop at all. She iterates a lot on the narration script with AI. Similarly, she spends a lot of time crafting each song. I'm really proud of her because she learned how to do this on her own and stitched everything together. https://t.co/vcr9KxPzXK

I've been waiting for years. Where are the AI products I was promised? https://t.co/BQufZdzzsu https://t.co/sdMzuO0YZQ

I've been waiting for years. Where are the AI products I was promised? https://t.co/BQufZdzzsu https://t.co/sdMzuO0YZQ

MotionV2V Editing Motion in a Video https://t.co/fxmYo1J4Te

discuss: https://t.co/EK4QLHdmEH

GLM-4.6V is now available in anycoder https://t.co/5vDWIyKNpw

app: https://t.co/esPDyHE1YC

Active Video Perception Iterative Evidence Seeking for Agentic Long Video Understanding https://t.co/lqW57WtpbR

discuss: https://t.co/ckKQIg9nqS

app: https://t.co/esPDyHE1YC

One-to-All Animation Alignment-Free Character Animation and Image Pose Transfer https://t.co/1BZPfKIPF6

discuss: https://t.co/rd3lGHU66F

model: https://t.co/zZmSBtQ11V

TwinFlow Realizing One-step Generation on Large Models with Self-adversarial Flows https://t.co/TZBE6hKBsv