Hello, AI Enthusiasts!
Welcome back to your weekly AI digest: human-curated, zero fluff.
Every Wednesday and Saturday, we bring you:
- The week's most important AI developments (no fluff)
- Hand-picked AI tools actually worth your time
- Expert insights on what these changes mean for you
Let's dive into this week's discoveries! ⚡
🔥 This Week in AI
⚡ Grok Imagine hits #1 on video leaderboards at 1/7th the price
👀 Bloomberg: Apple "runs on Anthropic" internally
🌍 Google opens AI world generator to the public
💰 Nvidia's $100B OpenAI investment plan stalls
🍎 Apple acquired Q.ai for $2B to read your lips
🎁 + 5 other news & articles you might like
🧰 +4 trending tools
xAI's Grok Imagine Dominates Video Leaderboards—At a Fraction of the Cost
xAI just released the Grok Imagine API, and it immediately jumped to #1 on Artificial Analysis rankings for both text-to-video and image-to-video generation—while undercutting competitors by up to 85% on price.
💰 Price Comparison (per minute with audio):
- Grok Imagine: $4.20
- Veo 3.1: $12.00 (3× more expensive)
- Sora 2 Pro: $30.00 (7× more expensive)
The API handles text-to-video, image-to-video, and video editing tasks with clips up to 15 seconds. The editing tools let you swap objects mid-scene, restyle entire environments, animate characters with custom performances, and shift backgrounds on command—all through natural language prompts.
Early partner integrations are already live through fal.ai, ComfyUI, InVideo, Flora, and HeyGen. HeyGen noted that Grok's model enables direct prompt-based edits for quick iterations, eliminating the need to regenerate entire clips when making small changes.
Why it matters:
This is the kind of price disruption that forces entire markets to restructure. If Grok Imagine maintains this quality at scale, the aggressive pricing could make it the default choice for creators and developers who need to iterate fast without burning budgets.
Elon Musk claims Grok Imagine is now generating more images and videos than all competitors combined. The video generation race just got significantly more competitive—and significantly cheaper.
Bloomberg Reveals: Apple "Runs on Anthropic" Internally
In a revelation that exposes the gap between Apple's public AI partnerships and internal reality, Bloomberg's Mark Gurman disclosed that Apple relies heavily on Anthropic's Claude models for its own product development—even as it partners with Google's Gemini for consumer-facing Siri features.
"Apple runs on Anthropic at this point. Anthropic is powering a lot of the stuff Apple's doing internally in terms of product development and a lot of their internal tools. They have custom versions of Claude running on their own servers internally, too."
— Mark Gurman, Bloomberg
The disclosure becomes even more interesting when you learn what almost happened: Apple initially planned to rebuild Siri around Claude. Negotiations fell apart when Anthropic demanded "several billion dollars a year" with prices doubling annually for three years. Apple deemed the terms excessive and went with Google instead—paying approximately $1 billion annually, a fraction of Anthropic's ask.
📊 The Deal That Didn't Happen:
| Provider | Annual Cost | Result |
|---|---|---|
| Anthropic (Claude) | $2-3B+ (doubling yearly) | ❌ Rejected |
| Google (Gemini) | ~$1B annually | ✅ Chosen |
But internally? Apple clearly values Claude enough to deploy it across product development, video processing, technical documentation, and engineering workflows. The company has negotiated custom Claude deployments running on Apple's own servers, suggesting data handling and privacy protections that justify the trust placed in external AI for proprietary work.
Why it matters:
This reveals how even Apple—famous for vertical integration and doing everything in-house—has determined that building competitive LLMs requires extraordinary capital investment that challenges even its financial reserves.
For Anthropic, this validates their enterprise strategy even after losing the consumer Siri deal. Powering Apple's internal operations at scale demonstrates Claude's capabilities in high-stakes production environments. The AI landscape isn't winner-take-all—it's specialized partnerships stacked on top of each other.
Nvidia's $100 Billion OpenAI Investment Plan Stalls
Nvidia's plan to invest up to $100 billion in OpenAI has stalled after internal doubts emerged about the transaction size and structure, according to a Wall Street Journal report.
📉 Deal Status:
| Original Plan: | $100B investment (non-binding) |
| New Discussion: | Tens of billions (equity round) |
| Jensen's Critique: | "Lack of discipline" in OpenAI's approach |
Nvidia CEO Jensen Huang has privately told industry partners the original agreement was non-binding and not finalized, while criticizing OpenAI's business approach and expressing concerns about competition from Google and Anthropic.
Why it matters:
This exposes potential cracks in what appeared to be one of AI's most powerful partnerships. The stall also highlights concerns about "circular financing"—Nvidia invests in OpenAI, which then buys Nvidia chips. Whether this is AI's "first domino to fall" or simply renegotiation depends on your perspective. But it's a reminder that even in the hottest sector of tech, fundamentals still matter.
Google Opens Project Genie: Create and Explore AI-Generated Worlds
Google DeepMind launched Project Genie, a web app that lets users create and explore AI-generated 3D worlds in real time. This comes five months after previewing the Genie 3 model that powers it.
✨ How it works:
Prompt a setting and character
Preview scene via Nano Banana Pro and Gemini
Navigate in first or third-person (walk, fly, drive)
Model remembers your world for consistent revisits
The technology operates at 720p resolution and 20-24 frames per second, generating the path ahead in real time. Sessions are currently capped at 60 seconds due to compute costs, with each user getting a dedicated "chip" while exploring.
720p
Resolution
20-24 FPS
Frame Rate
60 sec
Session Limit
🔒 Access Requirements:
Currently limited to Google AI Ultra tier ($250/month) subscribers in the US (18+). Access will expand to other tiers and international markets in the future.
Why it matters:
The applications are endless—robotics training, game prototyping, architecture walkthroughs, educational simulations, autonomous vehicle testing. We're watching the foundation being built for entirely new categories of software that don't exist yet.
✨ Tool Spotlight
Fireflies
Never take meeting notes again
If you've ever sat through back-to-back meetings only to realize you can't remember half of what was discussed or who committed to what, Fireflies solves that problem completely.
It's an AI meeting assistant that automatically transcribes and summarizes your meetings with powerful search, action item extraction, and team collaboration features—no manual note-taking required.
👥 Perfect for:
- Remote teams conducting frequent video meetings
- Project managers tracking decisions across calls
- Sales teams needing CRM integration
- Anyone who attends 5+ meetings per week
🎯 What makes it special:
Automatic transcription with speaker identification
Joins Zoom, Google Meet, or Teams calls automatically. Transcribes everything in real time with perfect speaker separation.
Intelligent extraction after meetings
Automatically pulls out action items, decisions, questions, and key topics. Search across all meetings like a Google Doc.
Seamless workflow integration
Auto-sync action items to Asana, Notion, or Slack. Send summaries to your CRM. Create custom tracking for sales or product discussions.
💰 Pricing:
Free plan with limited credits | Pro from $10/month | Business from $19/month
📌 What you should know:
- 90%+ accuracy for clear English audio
- Privacy note: Inform participants as it records meetings
- ROI justified for anyone spending 2+ hours weekly in meetings
⚡ Quick Hits
🍎 Apple's $2B bet on silent speech
Acquired Q.ai, an Israeli startup that reads facial micro-movements to interpret mouthed or whispered commands. Face ID creator's return to Apple. Could enable silent Siri interactions in AirPods/AR glasses.
Read more →⚖️ Anthropic sued by music companies for $3B+
Several music publishers allege unauthorized use of 20,000+ songs to train Claude. Lawsuit seeks over $3 billion in damages. Anthropic denies wrongdoing.
Read more →🚀 Blue Origin pauses space tourism for 2 years
Redirecting all resources to upcoming moon missions amid Trump administration pressure. New Shepard flew 38 missions carrying 98 humans past Kármán line.
Read more →🛰️ Amazon requests satellite deadline extension
Asks FCC for 2-year extension after admitting it will miss July 2026 deadline to deploy half its Project Leo constellation. Expected 700 satellites vs required 1,600.
Read more →🤖 AI agents get their own social network
Moltbook now lets AI assistants interact with each other, post to forums, and share information. OpenClaw has 100,000+ GitHub stars but serious security warnings.
Read more →🧠 Apple loses 4 AI researchers to rivals
Foundation Models team members departed for Meta and Google DeepMind, including senior Siri executive Stuart Bowers. Brain drain continues amid AI competition.
Read more →That's all for this edition! What tools are you most excited to try?
Reply to this email - we read every response and use your feedback to improve future editions.
See you on Saturday,
The AI Tool Discovery Team
Feedback
What'd you think of today's newsletter?
Vote below to let us know how we're doing.