Rand Xie

74 posts

Rand Xie

@Randxie29

Work on multimodal-LLM in xAI: Model, System, Product. Ex-Robinhood AI Lead, Ex-Googler

Katılım Ağustos 2016

813 Takip Edilen482 Takipçiler

Rand Xie@Randxie29·4 Haz

Thanks to Jack for the shoutout! Large projects have the same credit assignment problem as long-horizon agents: delayed rewards and long causal chains make attribution hard. Maybe we need an ensemble of value models. 😄

Jack Cai@JackCaiXun

Unforgettable time! Grok Imagine was a true collective effort. In addition, @JiachenLi11 @chaoqi_w @jefffhj @Randxie29 — and many others made significant contributions in the initial 3 months sprint

English

3.5K

Rand Xie retweetledi

Xuhui Jia@jia_xuhui·31 May

Grok Imagine is getting better and better. Our only goal is to make it genuinely useful. If we do that well, strong rankings will naturally follow as a byproduct.

Arena.ai@arena

Grok-Imagine-Video-1.5-Preview (720p) has landed #1 in the Image-to-Video Arena! This is a massive +52 pt improvement over Grok-Imagine-Video (720p), surpassing the best video models Seedance-2.0 and HappyHorse. Congrats to @xAI and @elonmusk on this big achievement!

English

191

27.7K

Rand Xie retweetledi

Guodong Zhang@Guodzh·4 Haz

@EthanHe_42 worked with me and @imhaotian for half a year on Grok Imagine. I don't think he was being intentional overclaiming here, but the internet narrative quoting him as the "lead" was werid. It was really driven by @imhaotian most of the time. Good times — especially that intense three-month sprint at the start. Huge credit goes to @imhaotian and several other key people like @zeliu_ @jathushan @ZhibeiM @hexiang @JackCaiXun @chaitu, and later @jia_xuhui @YknZhu, along with many others — especially the latest Grok Imagine 1.5.

Ethan He@EthanHe_42

In @latentspacepod podcast, I shared my view on video generation, world models, LLMs, agents, continual learning and where the next frontier is. 1. Video models get most of their intelligence from language, not from video data. 2. Idea-to-code is fast now. The bottleneck is back to having enough compute to try every idea. 3. Iteration speed beats almost everything else in model development. 4. The next leap won't be a better video model. It'll be a video agent. 5. Diffusion will be the frontend of AGI, the LLM the backend. Generative UI will replace HTML/CSS: user intent straight to pixels. 6. Physical embodiment may become a tool a powerful AI picks up. Robotics may get solved by video-capable LLMs. 7. Continual learning may look like models that manage their own context, and even rewrite their own harness at test time. Thanks @swyx and @vibhuuuus for having me 🙏 youtube.com/watch?v=jPtQlI…

English

304

98.1K

Rand Xie@Randxie29·24 May

A question I keep asking myself: Do you want to ride the waves and capture the upside everyone already sees? Or stand with a small group of people doing deeply contrarian things before the world understands why they matter?

English

435

Rand Xie@Randxie29·9 Nis

@HeinrichKuttler @xai @elonmusk Thanks so much for setting up all those GPUs, Heiner! And huge congratulations on the new family member!

English

186

heiner@HeinrichKuttler·9 Nis

Update on the below: 9 months later we're expecting our third child any day now. I decided to focus on my family for a while. Leading Supercomputing at @xai will remain the honor of my life. Thank you @elonmusk for the opportunity. We will win.

heiner@HeinrichKuttler

me trying for kid no. 3:

English

553

46.9K

Rand Xie retweetledi

Guodong Zhang@Guodzh·3 Nis

Surprised again and again how much a few engineers can do with a clear goal in 3-6 months.

English

142

14.3K

Rand Xie retweetledi

Elon Musk@elonmusk·31 Mar

@xdNiBoR The future of AI is primarily video understanding and generation, because photons are by far the highest bandwidth form of communication. These are essential tools for AGI. Worth mentioning that Imagine is positive gross margin for @xAI, not a money loser.

English

382

575

7.7K

1.4M

Rand Xie@Randxie29·29 Mar

Grok Imagine is often overlooked in conversations around video generation, but the market share speaks for itself. It’s been incredibly rewarding to take it from nothing to where it is today. There are many stories behind this journey that are hard to fully capture here, but every sleepless night, pressure, and uncertainty has been worth it.

English

226

Rand Xie retweetledi

X Freeze@XFreeze·27 Mar

The speed at which Grok has captured the AI video market over the last few months is wild Grok is completely taking over…..users are already generating way more videos with Grok Imagine than on any other AI platform The traffic shift is unreal. Grok’s dominance moved so fast it crushed demand entirely, shutting down models like Sora Adoption rate is insane. xAI is moving at lightspeed If you haven’t tried Grok Imagine yet, you’re missing the new standard 🔥

English

154

6.4K

Rand Xie@Randxie29·25 Mar

@karpathy I think the problem isn't memory, it's passive memory. We need agentic memory that updates and prunes through interactions. Humans don't remember every detail about their friends - they forget about noises, and remember those important moments.

English

102

Andrej Karpathy@karpathy·25 Mar

One common issue with personalization in all LLMs is how distracting memory seems to be for the models. A single question from 2 months ago about some topic can keep coming up as some kind of a deep interest of mine with undue mentions in perpetuity. Some kind of trying too hard.

English

1.8K

1.1K

21.1K

2.8M

Rand Xie@Randxie29·22 Mar

@alilbitofmo @xai I didn’t get a chance to say goodbye in person. You’ll be missed. It's been a pleasure working with you on all the human data tasks!

English

167

alilbitmo@alilbitofmo·21 Mar

Today was my last day at @xai . The past year (and some change) have been nothing short of exceptional—inspiring and grueling in equal measure. It was an honor to support the Imagine team from the human data side, exhilarating to champion and deliver impossible outcomes on impossible timelines, and deeply inspiring to see our work improve the model in real time. If I didn't get to thank you personally today, here's my heartfelt thanks to the Omni/Imagine team, the Image & Video Human Data crew, and all my colleagues who slogged through the trenches with me day after day (across eng, HD, and beyond). It was a true honor working alongside you—I’ll miss you all very much. Excited for whatever comes next, and rooting for the xAI team! 🚀 Ad astra, etc.

English

4.3K

Rand Xie retweetledi

Yukun@YknZhu·13 Mar

Give it a try, it is fun!

Grok@grok

Create a video using up to 7 images in Grok Imagine. Include characters, places, objects, and more. Available on iOS, Android, and Web!

English

5.4K

Rand Xie retweetledi

Déborah@dvorahfr·2 Mar

Example of using the Grok Imagine extension: Here is a 26-second scene without cuts, without editing, perfect stylistic coherence, made in 5 minutes. I requested two 10-second extensions and one 6-second extension. There is no loss of quality throughout the sequence and perfect sound consistency.

English

390

345

2.6K

16.4M

Rand Xie retweetledi

Elon Musk@elonmusk·2 Mar

Extend your video to 30 seconds

X Freeze@XFreeze

Grok Imagine just got a cool new feature You can now extend your videos up to 30 seconds with the brand-new Frame Extend option Generate a video → pick any moment in the timeline and keep going

English

4.7K

5.8K

47.5K

22.1M

Rand Xie retweetledi

Elon Musk@elonmusk·1 Mar

Grok Imagine

Déborah@dvorahfr

Grok Imagine UPDATE: Once your animation is complete, you can extend it using any frame you like. This is really useful. Another great feature is that the music continues when the animation is extended. In the example below, I was able to extend the animation to 30 seconds without any variation in style or distortion. Look in your video options, you will see video extension appear.

English

3.9K

5.4K

34.3K

12.8M

Rand Xie@Randxie29·19 Şub

Great article! "The scarce thing flipped from execution to judgment: can you orchestrate systems, run parallel bets, and have the taste to know which results matter?"

Amy Tam@amytam01

x.com/i/article/2023…

English

209

Rand Xie@Randxie29·17 Şub

New tool will bring a new art form. This short film reminded me of my experience as mechanical engineer: everyone was trained with new CAD tool (no more manual drawing). Tools improve efficiency, not creativity. Those who can create generates much better 3D design.

银河百科全书@yhbkqs

贾樟柯和豆包视频生成模型 Seedance 2.0合作，共同完成了这支有些特别的短片《贾科长Dance》。屏幕里的两个“贾樟柯”，都是通过Seedance 2.0生成的。其中一个保留了明显的“AI感”，另一个，几乎就是现实生活中的贾樟柯。人类已经无法阻止AI的进步。要思考的是，AI能给人类带来什么？ t.cn/AXto0PIT

English

289

Rand Xie@Randxie29·15 Şub

Why can films tell stories? Camera movement, lighting, and acting all matter - but montage is the decisive element. Narrative lives between shots. Even the simplest story requires more than a single shot; without montage, cinema is nothing more than a surveillance camera left running.

English

Rand Xie@Randxie29·13 Şub

@hangg70 You will be missed! I still remember the time we walked together near the office, and discuss the meaning of writing. Best wishes to your next journey!

English

206

Hang Gao@hangg70·11 Şub

I left xAI today. It was truly rewarding to contribute to grok imagine video series: 0.9 as our first release, then 1.0 that recently topped across competitive leaderboards and user feedback. I see a mix of humble craftsmanship and ambitious vision throughout the team. They taught me about what I want and how I want to proceed in my career. Thank you to everyone who made this journey unique and memorable.

English

192

106

430.6K

Rand Xie@Randxie29·13 Şub

@jefffhj Love the word "soul"! Taste becomes much more important when generated video gets longer.

English