There seems to be some confusion about X Premium+ and what perks it includes! To be clear: X Premium+ *includes* SuperGrok. If you want to use the latest @Grok features like Grok Imagine and Grok 4.20, it’s a better deal to get X Premium+ instead of just SuperGrok. https://t.co/D4nNkZgAbF

Grok Imagines generates concert videos and original music https://t.co/lJYjvd2UuA

wow.. new Grok Imagine can keep 7 things incredibly consistent, check prompt below https://t.co/8SAb88Opsn

Try Grok, now with the ability to combine your images into one unique creative video or animation — absolutely fantastic and so easy to use! https://t.co/pzOaTClYal

One of the largest barriers to saving our nations is that most Westerners genuinely struggle to believe people like this exist. We fish around for explanations because it’s just incomprehensible. We can’t accept the vastness of human differences. https://t.co/jLzpyLKBCz

Demographic diversity is a way to mask ideological conformity https://t.co/m2qhv70Flg

Grok Imagine offers a wider range of tools: image animation image editor video extension And now Up to 5 image reference https://t.co/QJdjPpzK7q

Today is World Sleep Day. https://t.co/8LebF8nfBC

Today is World Sleep Day. https://t.co/8LebF8nfBC

xAI's Grok Imagine just took #1 on the Video Arena leaderboard (DesignArena by Arcada Labs) → Elo: 1336 - 33 points clear of #2 → Win rate: 69.7% - highest on the entire board → Battles: 15,590 - statistically rock solid Beats Veo 3 Fast, Veo 3.1 Fast, and Veo 3.1 - all at once And it does it in 21.3 seconds, while Google's models take over 40 seconds xAI went from nowhere to #1 in AI video. The pace of Grok Imagine improvement is unmatched

This is this Claude's first knowledge of Codex but yet "Classic" Codex collateral damage hahahahaha https://t.co/LTY3ACBvmI

ai engineer is coming singapore in may and we're looking for kickass speakers - tag anyone you think would be a great fit for these tracks: - software: learn how builders are shipping, scaling, and understanding AI applications in production - design: discover how designers are moving beyond chat to create the next generation of intuitive, non-deterministic interfaces - robotics: see how foundation models are crossing into the physical world @agrimsingh @ivanleomk @aimuggle @swyx

Bookmark RAG agent in @CorpusAIapp powered by the new Qwen3.5 2B 4-bit model. Thinking mode is a little slow but producing good summaries and tool calls so far 👍 https://t.co/XojR1HhjGa

Today, Genspark launched Genspark Claw, powered by Genspark Cloud Computer, and we surpassed $200M in run rate in just 11 months, doubling in the last two months alone. Genspark Claw introduces a new execution layer to the AI workspace: a dedicated cloud computer per user, enabling AI agents to operate software environments on your behalf with privacy by isolation and clear permissioning. We also closed our Series B extension round, led by Emergence Capital—making us the largest investment in Emergence Capital’s history. Investors in this round, alongside Emergence, include: Japan SBI, Korea Mirae Asset, HartBeat Ventures by Kevin Hart, Markham Valley Ventures by Simu Liu and Keisuke Honda. As part of this release, we’re rolling out updates across the AI Workspace 3.0 suite to support more hands-free execution and collaboration: - Genspark Workflows - Genspark Teams (a Slack rewrite and it is FREE!) - Genspark Meeting Bot - Speakly for iOS and Android - Genspark Chrome Extension - Genspark Realtime Voice

Grok 4.20 Multi-Agent just took #1 on BridgeBench. 96.1 overall. 100% completion rate. 87.8s latency. Beat GPT 5.4, Claude Sonnet 4.6, and Claude Opus 4.6. xAI came out of nowhere. The multi-agent future is here. And Grok is leading it. https://t.co/D28gT6v1CW

The backlash against MCP isn't a surprise for readers of Understanding AI. @startupandrew predicted it in December. https://t.co/PzaVq5r5cv

The answer is simple: The world doesn’t need another AI researcher. The world needs the greatest AI educator, thinker, and taste-maker - Andrej Karpathy. https://t.co/PmaSCB8iXd

@AAAzzam @HanchungLee Also you know someone gonna take this seriously https://t.co/C5efq2wKhl

next-gen 💻 🫧☄️ https://t.co/X7Z8XfEpbQ

next-gen 💻 🫧☄️ https://t.co/X7Z8XfEpbQ

I’m genuinely in tears right now please unmute this video https://t.co/WthJDcmov0

I’m genuinely in tears right now please unmute this video https://t.co/WthJDcmov0

🚨: There have been thousands of generations of humans, and you are alive to witness the first photo of a Sunset on another World.😮 This is a real photo of the sunset on Mars. https://t.co/jmZSO7zvrB

https://t.co/u4mN4Xco0W

https://t.co/u4mN4Xco0W

@Viggie_Smalls93 https://t.co/QMr86vkhDm

In March 2003, Jack Nicholson, who had recently received his 12th Oscar nomination for “About Schmidt,” invited his fellow best actor contenders to his home on Mulholland Drive. The U.S. had just invaded Iraq, and he wanted the other men — a group that included past winners Daniel Day-Lewis, Michael Caine and Nicolas Cage — to boycott the ceremony with him. But Adrien Brody, the only nominee without an Academy Award of his own, wasn’t having it. “I said, ‘I don’t know about you guys, but I’m going,’” Brody recounted to The Sunday Times in 2023. “I said, ‘I kind of have to show up. My parents are coming. This doesn’t come around too often. I know you guys are all winners. You can sit it out. But I can’t.’” Brody’s mind may have been made up, but the war plunged Hollywood into a state of uncertainty as the town braced for the 75th Academy Awards. Read "Looking Back at the 2003 Oscars" here: https://t.co/erG5RTk8AQ

自衛隊が前線の情報分析にＡＩ活用へ…敵の写真や音声を自動解析し統合、配置の立案も : 読売新聞オンライン https://t.co/rMMNtdbJYQ 　＃自衛隊ＡＩ活用　＃敵の写真や音声を自動解析

自衛隊が前線の情報分析にＡＩ活用へ…敵の写真や音声を自動解析し統合、配置の立案も : 読売新聞オンライン https://t.co/rMMNtdbJYQ 　＃自衛隊ＡＩ活用　＃敵の写真や音声を自動解析

🎉 Congrats to @nvidia on the release of Nemotron 3 Super — day-0 support in vLLM v0.17.1! Verified on NVIDIA GPUs. 120B hybrid MoE, only 12B active at inference. Big upgrades over the previous Nemotron Super: - 5x higher throughput - 2x higher accuracy on Artificial Analysis Intelligence Index - Multi-Token Prediction (MTP) for faster long-form generation - Configurable thinking budget — dial accuracy vs token cost per task - 1M token context window Supports BF16, FP8, and NVFP4. Fully open: weights, datasets, recipes. Blog: https://t.co/PAN0y778iB 🤝 Thanks @NVIDIAAIDev Nemotron team and vLLM community contributors!

Interesting new paper on online RL for agents. Most agent training still treats deployment and learning as separate phases. Serve the model first, collect data later, fine-tune offline. But every agent interaction already contains a learning signal. This paper introduces OpenClaw-RL, a framework that trains agents from the next state that follows each action: user replies, tool outputs, terminal traces, GUI changes, and test results. The key idea is to recover two signals at once. Evaluative signals become scalar rewards through a PRM judge. Directive signals become token-level supervision through hindsight-guided on-policy distillation. In their personalization setup, the combined method improves the score from 0.17 to 0.81 after 16 update steps, outperforming binary RL or OPD alone. They also show gains across tool-call and GUI agents when combining process and outcome rewards. Why it matters? Agents should improve simply by being used. This is a practical step toward online, always-learning agent systems instead of static deployments. Paper: https://t.co/bzLdREoet8 Learn to build effective AI agents in our academy: https://t.co/LRnpZN7L4c