Good morning, AI enthusiasts. Jeff Bezos is back in an operational CEO role with $6.2 billion in backing for Project Prometheus, his physical-world AI startup poaching talent from OpenAI, DeepMind, and Meta.
The funding dwarfs most competitors and signals Bezos believes manufacturing and engineering AI represents the next frontier beyond chatbots.
In today's recap:
Bezos launches $6.2B Project Prometheus for physical AI
LeCun exits Meta to build world models startup
Google debuts Nano Banana Pro with 4K generation
8 new AI tools and more
BEZOS & MANUFACTURING
Bezos returns as co-CEO for AI startup
Source: Getty Images
Recaply: Jeff Bezos just launched Project Prometheus with $6.2 billion in funding, according to The New York Times, marking his first operational CEO role since leaving Amazon in 2021.
Key details:
Project Prometheus focuses on AI for engineering and manufacturing across computers, aerospace, and automobiles, using models that learn from physical experiments rather than text analysis alone.
The $6.2 billion backing makes it one of the most well-financed early-stage startups globally, dwarfing Periodic Labs' $300 million and exceeding even Thinking Machines Lab's $2 billion raised this year.
The New York Times reports nearly 100 employees have been hired from OpenAI, DeepMind, and Meta, with Bezos' co-CEO Vik Bajaj bringing experience from Google X projects like Waymo and Wing.
The company has kept a low profile since launch, with its founding date unclear and base location still undetermined as operations ramp up.
Why it matters: Bezos' return to an operational CEO role signals his belief that physical-world AI represents the next major frontier beyond chatbots. The crowded market already includes Google, Meta, OpenAI, and Anthropic racing toward similar breakthroughs, but Project Prometheus' massive funding and focus on manufacturing positions it as a direct challenge to incumbents. Whether Bezos can replicate his Amazon success in AI, especially competing against the very companies he's poaching talent from, remains the central question.
AI RESEARCH
LeCun leaves Meta for world models startup
Source: Yann LeCun
Recaply: Meta chief AI scientist Yann LeCun is leaving to launch a startup focused on world models and Advanced Machine Intelligence, according to CNBC, with Meta partnering on the venture despite ongoing tension within its AI division.
Key details:
LeCun's startup will pursue AI systems that understand the physical world, have persistent memory, can reason, and plan complex action sequences, moving beyond current large language model approaches.
The 65-year-old joined Facebook in 2013 to create the FAIR division, with his 12-year tenure spanning Meta's entire modern AI era from research lab to commercial competition.
People familiar with the matter said LeCun rarely interacted with new chief AI officer Alexandr Wang or the TBD Labs unit, revealing deep divisions between Meta's research-focused old guard and product-focused new leadership.
Meta laid off 600 positions from its Superintelligence Labs division in October, with the cuts hitting FAIR researchers before LeCun's departure announcement this week.
Why it matters: This coming departure is not a surprise. There has been tension between Meta's old guard and new hires throughout the reorg, and LeCun's AI critiques are constantly at odds with Meta's new AI direction. Time will tell if the upcoming venture ultimately proves him right or if the high-profile talent proves its worth instead.
Google's Nano Banana Pro goes 4K
Source: Google
Recaply: Google just released Nano Banana Pro (Gemini 3 Pro Image), a 4K-capable image model with grounding via Google Search and integration across Adobe, Figma, and Google's new Antigravity IDE.
Key details:
The model gives professionals control over lighting, camera angles, depth of field, focus, and color grading to ensure studio-quality outputs with 2K or 4K resolution.
Nano Banana Pro costs $0.139 per 2K image and $0.24 per 4K image, roughly 3.5x and 6x more expensive than the original Nano Banana's $0.039 per 1024px image.
Google integrated SynthID watermarking directly into every generated or edited image, with plans to add C2PA content credential detection for verification over time.
The model rolls out in paid preview through Gemini API, AI Studio, and Vertex AI, while Gemini app users on free tiers get limited generations before defaulting to Nano Banana.
Why it matters: Google is making a clear bet that professionals will pay premium prices for production-ready image generation. Adobe and Figma integrations signal that AI image tools are moving from experimental to essential in creative workflows. With SynthID baked in and C2PA coming, Google is positioning itself as the ethical choice in a crowded field where provenance matters.
NEWS
Beyond the big stories in AI today
AI Godmother Fei-Fei Li urges young AI talent to focus on passion and mission over minute career details, according to Business Insider.
Google DeepMind released WeatherNext 2, generating hundreds of weather scenarios in under one minute with 99.9% improved accuracy over its predecessor.
OpenAI fixed ChatGPT's em dash overuse, allowing users to disable the punctuation through custom instructions in personalization settings.
Anthropic invests $50B in U.S. data centers across Texas and New York, creating 800 permanent jobs and 2,400 construction jobs.
OpenAI pushes back against New York Times demand for 20M private ChatGPT conversations in ongoing copyright lawsuit.
World Labs launched Marble, a multimodal AI world model creating 3D environments from text, images, or video with interactive editing.
Perplexity launched Comet AI browser on Android with built-in ad blocker, voice mode, and tab summarization capabilities.
White House drafts executive order establishing AI Litigation Task Force to challenge state AI laws and withhold $42B in federal funding, according to CNBC.
OpenAI rolled out group chats globally in ChatGPT, allowing up to 20 people to collaborate in shared conversations across all plans.
TOOLS
Trending AI Tools
📚 Talkio - AI voice tutors for practice and pronunciation in 40+ languages
🚀 Nuelink - Automate social media scheduling across all platforms
🎤 FineVoice - Clone any voice and create AI speech in 1500+ styles
🔊 Krisp - Remove background noise, transcribe meetings, convert accents in real time
🎨 PromeAI - Turn design sketches into realistic architectural renderings
📝 Clipto - Convert audio and video to text with AI transcription
🎬 Arcads - Generate ad-style videos with AI actors and visuals
⚡ PopAI - Create slides, outlines, and visuals from any topic instantly
