| |
Good morning, {{ first_name | AI enthusiasts }}. xAI is officially rolling out Grok Imagine, its new tool for ultra-fast AI video generation with native sound.
By claiming to produce videos in just half to a quarter of the time it takes to generate AI images, the company is pushing the speed limits of creative AI. But the big question remains: can it match the cinematic quality of Google’s leading Veo 3 model?
In today’s AI rundown:
-
xAI rolls out Grok Imagine AI video generator
-
Google releases Gemini 2.5 Deep Think
-
How to extend AI videos beyond 8 seconds
-
Study: Anthropic looks into AI’s personality shift
-
4 new AI tools & 4 job opportunities
LATEST DEVELOPMENTS
OPENAI
🎥
Image source: xAI / Screenshot via @omooretweets on X
The Rundown: xAI is officially Grok Imagine — its new AI image and video generator — to all SuperGrok and Premium+ X subscribers on its iOS app, entering a space dominated by Google, OpenAI, Runway, and China’s viral generators.
The details:
-
With Grok Imagine, users can turn any simple text prompt or image into a 15-second video, complete with native audio, in seconds.
-
The model even creates images from text, which can eventually be animated into stylized videos, and continues to as users scroll down.
-
In terms of quality, the videos generated by Imagine look more “AI-generated” as compared to other video models, delivering cinematic outputs.
-
Elon Musk the model “should get better every day” while that it makes videos in 1/2 to 1/4 the time rivals take to make a single image.
Why it matters: While Grok Imagine doesn’t appear to outperform other top video generators on the market, xAI is aiming to offer a fresh take on video generation with it’s playful and unfiltered style. But as we’ve seen with the , that approach can sometimes lead to unexpected (and even controversial) results.
TOGETHER WITH AMAZON
⭐️
The Rundown: The global scale and scope of Amazon gives AI experts more paths to pursue artificial intelligence that matters. From leading-edge research to real-world impact, Amazon is always seeking new ways to make people’s lives better with AI.
Watch the video to explore the advantages of working at Amazon, including:
-
Industry-leading infrastructure, custom silicon chips, and global customers
-
1,000+ AI services and applications in play or in progress
-
A team of AI industry leaders, solving real problems for real people
-
The array of ways you can pursue your passion and grow your AI career
🧠
Image source: Google
The Rundown: Google Gemini 2.5 Deep Think, its first publicly available multi-agent model that does “parallel thinking” to help researchers, scientists, and academics tackle complex problems.
The details:
-
First announced at I/O 2025, Gemini 2.5 Deep Think is a variant of the model that won the at this year’s International Math Olympiad.
-
When handling hard questions, the model spawns multiple agents to explore possible solutions in parallel and then decides the best answer from them.
-
It scored 34.8% on Humanity’s Last Exam, surpassing Grok 4 and OpenAI’s o3, while delivering SOTA performance on coding and .
-
Gemini 2.5 Deep Think is rolling out to Gemini app users on Google’s $250/month Ultra plan, with the IMO variant accessible to select researchers.
Why it matters: While Meta is vying for “personal” superintelligence, Google is taking a different route — empowering researchers, scientists, and academics with a parallel-thinking AI that, instead of offering direct answers, spawns a team of expert minds to tackle problems from multiple angles before converging on a solution.
AI TRAINING
🎬
The Rundown: In this tutorial, you’ll learn how to bypass Google Veo 3’s 8-second clip limit using Google Flow’s frame-to-video feature that lets you create longer videos with character consistency.
Step-by-step:
-
Create your initial 8-second video using “Text to Video” in (choose Veo 3 for audio)
-
Open the scene editor and save the final frame using “Save frame as asset”
-
Start a new “Frames to Video” generation using your saved frame, and describe the next sequence
-
Combine both clips in the timeline editor and trim for smooth transitions
Pro tip: Repeat this process multiple times to create even longer videos while maintaining perfect character consistency throughout.
PRESENTED BY CONCIERGE
🪄
The Rundown: Wish AI understood your work context better? Concierge solves this by connecting your software stack (ex: Gmail, Linear, Slack, Attio, Jira, Confluence) seamlessly to all your favorite AI models (GPT, Claude, Grok, Deepseek, Perplexity).
With Concierge, you can:
-
Instantly integrate your apps and get real-time answers based on your personal data
-
Choose from all the top AI models without juggling multiple subscriptions
-
Designed for Product Managers and Product Marketers at modern companies
and get AI to truly work for you.
AI RESEARCH
😈
Image source: Anthropic
The Rundown: Researchers at Anthropic just “Persona Vectors,” neural network activations that help understand and control unexpected (sometimes even unsettling) behavioral changes demonstrated by AI models.
The details:
-
While trained to be helpful and honest, AI models can sometimes drift away, exhibiting unexpected personality traits like sycophancy or racism.
-
When these behavioral changes happen, certain patterns of activity or persona vectors are seen within an AI’s neural network, like the human brain.
-
Researchers extracted these vectors by comparing activation patterns between opposing behaviors (evil vs non-evil).
-
They focused on three traits—evil, sycophancy, and hallucination—using persona vectors to reduce their emergence and narrow down causative data.
Why it matters: With popular AI tools like and previously showing behaviors such as sycophancy and antisemitism, it’s clear that no model is immune to behavioral drift. Anthropic’s research offers a promising path to understanding these shifts at the neural network level—and using that understanding to build safeguards.
QUICK HITS
🛠️
-
📺 – Create a video course from any blog, PDF, or URL
-
⚙️ – First AI agent built for the MCP ecosystem
-
📽️ – Runway’s video editing model, now available via API
-
👁️ – Cohere’s 112B VLM with enterprise security
💼
-
🚀 – AI Educator (Workshop Host)
-
🎓 – Business Development Lead
-
🔍 – Corporate Finance Manager
-
⚙️ – Software Engineer
📰
European AI startup Mistral is looking to raise $1B at a $10B valuation from multiple VCs and Abu Dhabi’s MGX as the AI race heats up.
OpenAI an opt-in feature in ChatGPT that allowed users to make their conversations discoverable by search engines, such as Google.
Anthropic OpenAI’s access to its API over violation of terms of service and for the heavy usage of Claude Code among OAI tech staff ahead of GPT-5’s release.
Apple has formed an “Answers, Knowledge, and Information” team to create a ChatGPT-like app that can respond to queries using information from the web.
Apple’s CEO, Tim Cook, also analysts that the iPhone maker is “open to M&A” that accelerates its AI roadmap and helps catch up to rivals.
Amazon CEO Andy Jassy that the company’s new AI-powered assistant, Alexa+, may eventually deliver ads to users during conversations.
Meta is aiming to $2B worth of data center assets to outside partners as it works to set up massive data centers to power its superintelligence mission.
COMMUNITY
🎥
Check out our last live workshop with Dr. Alvaro Cintas, The Rundown’s AI professor. By the end of this workshop, you’ll have practical strategies to get the AI to do exactly what you want.
Watch it . Not a member? Join on a 14-day free trial.
See you soon,
Rowan, Joey, Zach, Shubham, and Jennifer—the human’s behind The Rundown
