Welcome back!

OpenAI is preparing for a $1 trillion IPO as soon as next year, having completed its restructuring into a full for-profit while keeping its nonprofit foundation in control. With Microsoft holding a 27% stake, this could deepen the links between AI and global finance. Let’s unpack…

Today’s Summary:

• 🏦 OpenAI prepares $1 trillion IPO

• 🐶 OpenAI Sora adds pets to videos

• 🧠 Andrej Karpathy says AGI a decade away

• 💡 Signs of introspection in language models

• 🎵 OpenAI developing new music AI

• 🗂️ Claude Memory for Pro & Max

• 🛠️ 2 new tools

TOP STORY

OpenAI Prepares $1 Trillion IPO

The Summary: OpenAI has restructured into a full for-profit company, controlled by a nonprofit foundation, ending the profit cap that once limited investor returns. Microsoft holds a 27% stake. OpenAI is preparing for an IPO that could value it at up to $1 trillion by 2026 or 2027. At that valuation, it would rank among the twelve largest publicly traded companies in the world.

Key details:

• The OpenAI Foundation now controls the for-profit OpenAI Group

• Microsoft holds 27% and retains API exclusivity through 2032

• OpenAI’s revenue run rate nears $20 billion, with $7 billion losses

• The IPO could raise roughly $60 billion, one of the largest in history, enabling OpenAI to fund its compute infrastructure

Why it matters: AI is becoming inseparable from global capital flows, as data centers become the new oil fields, geopolitical and deeply financial. OpenAI’s IPO may institutionalize AI as a new asset class that competes with energy and sovereign bonds.

FROM OUR PARTNERS

The AI That Remembers Everything You Say

The AI Wearable That Makes Your Life Unforgettable

Your greatest asset is your time. So stop wasting it jotting notes or chasing forgotten conversations.

The Limitless AI Pendant captures, transcribes, and organizes every meaningful interaction automatically. Get instantly searchable transcripts, smart summaries, and contextual reminders - all at your fingertips, all fully encrypted.

Tap into the future of productivity and free your mind to focus on what truly matters with Limitless.

Learn More

OPENAI

OpenAI Sora lets you add your pet to AI videos

The Summary: OpenAI’s Sora app now lets users turn pets, toys, or drawings into reusable “character cameos” for AI videos. These characters can be shared, named, and reused with permission. The update also adds video stitching for multi-scene generation, and leaderboards.

Key details:

• Character cameos capture short clips of pets, objects, or AI-generated figures that can appear in future videos

• Users can set permissions as Only me, Mutuals, or Everyone, and revoke access anytime

• New clip stitching enables longer narratives and storytelling

• Paid “extra gens” launch alongside plans for a Sora economy where creators can charge for cameo use

Why it matters: Adding pets to AI videos may look playful, but it subtly tests a new structure for digital ownership. Sora treats characters as portable identity units that can be reused, licensed, and monetized. The platform could evolve around user-defined assets that persist beyond any single video, drawing a blueprint for identity-based media.

FROM OUR PARTNERS

100 Ways to Work Smarter with AI

Want to get the most out of ChatGPT?

ChatGPT is a superpower if you know how to use it correctly.

Discover how HubSpot's guide to AI can elevate both your productivity and creativity to get more things done.

Learn to automate tasks, enhance decision-making, and foster innovation with the power of AI.

Download the free guide.

INTERVIEW

Andrej Karpathy says AGI is a decade away

The Summary: In a wide-ranging interview, Andrej Karpathy, OpenAI cofounder and former Tesla AI director, gave a reality check on the race toward artificial general intelligence. He says the industry’s focus on AI agents is far ahead of what current systems can do, with most still unreliable. Karpathy expects it will take a decade of hard work before true agentic intelligence appears.

Key details:

• Karpathy calls this “the decade of agents”, saying AGI will take longer than expected

• Reinforcement learning should be replaced by richer methods

• Current agent systems often produce error-prone code

• He contrasts “animals”, which are embodied and evolved, with “ghosts” LLMs packed with static knowledge

• Real progress will come from models that can build and update internal world models

Why it matters: Karpathy identifies a key bottleneck: AI correlates but doesn’t yet understand. The next step is to build systems that can form and refine internal world models. Moving from 99% to 99.999% reliability demands a full rethink of how cognition works. That’s the decade-long grind he’s describing.

QUICK NEWS

Quick news

• Signs of introspection in large language models

• OpenAI is developing a music AI model

• Claude Memory available to Pro and Max users

TOOLS

🥇 New tools

• Cursor 2.0 - With new composer and multiagents interface

• Alai - Create high quality presentations with AI