Find out why 1M+ professionals read Superhuman AI daily.

In 2 years you will be working for AI

Or an AI will be working for you

Here's how you can future-proof yourself:

1. Join the Superhuman AI newsletter – read by 1M+ people at top companies

2. Master AI tools, tutorials, and news in just 3 minutes a day

3. Become 10X more productive using AI

Join 1,000,000+ pros at companies like Google, Meta, and Amazon that are using AI to get ahead.

Sign up and start learning AI

Today:

• Ex-Google Insiders Decode Meta’s $14B Scale AI Move, Apple’s AI Doubts, and the Rise of Self-Improving Models

• Midjourney Launches AI Video Tool

• Meta Tried Buying Ilya’s Startup

• MiniMax Launches Hailuo 02 Model

• Meta Eyes Friedman, Gross, NFDG

ex-Google Director Just Revealed What's Coming Next…

Meta is acquiring a major stake in Scale AI for $14 billion, aiming to revamp its AI push after LLaMA 4's weak reception. The deal brings Scale’s founder, Alexander Wang, to lead Meta’s new superintelligence efforts. Critics question the high price and strategic value as synthetic data rises and traditional labeling slows. 

Meanwhile, debates grow over AI reasoning, model self-improvement, and whether recursive self-training signals an oncoming intelligence explosion.

WATCH THE VIDEO ON YOUTUBE

Midjourney Launches AI Video Tool

Midjourney has launched a tool that turns still pictures into short videos. After creating or uploading an image, click “animate,” describe the motion, and get a five-second clip. You can extend it four times to 21 seconds and choose motion. The feature works on web and Discord and needs a subscription—a paid monthly plan—costing roughly one picture per video second. Disney and Universal are already suing over alleged copyright violations.

Why this matters

1. Creative leap – Moving from synthetic images to moving pictures shows rapid progress in text-to-video, bringing Hollywood-style animation closer to everyday users.
   

2. Cost signal – Midjourney’s price (about one image per video second) gives a public yardstick for the heavy computing bills behind generative video.
   

3. Legal spotlight – Disney and Universal’s lawsuit highlights looming rules on how AI systems learn from copyrighted movies, a fight that could shape future model training and access.

Meta Tried Buying Ilya’s Startup

Meta tried to buy Safe Superintelligence, the $32 billion AI firm started by OpenAI co-founder Ilya Sutskever. After Sutskever rejected the offer, CEO Mark Zuckerberg began recruiting SSI’s co-founder Daniel Gross plus venture partner Nat Friedman. Both will join Meta under Scale AI veteran Alexandr Wang, while Meta buys a share of their VC fund NFDG. The moves deepen Meta’s costly push to secure top engineers for artificial general intelligence research aims.

Why this matters

1. Talent arms race – Meta’s willingness to spend billions and even poach fund partners shows how fierce competition for elite AI researchers has become.
   

2. Safety vs. scale tension – Safe Superintelligence was founded to pursue “safe” advanced AI; Meta’s pursuit hints at big tech’s desire to fold safety-oriented startups into massive, profit-driven labs.
   

3. Funding signal – A $32 billion valuation for a one-year-old startup underscores how capital is flooding into AGI efforts, shaping which approaches and players will dominate.

MiniMax Launches Hailuo 02 Model

Chinese startup MiniMax unveiled Hailuo 02, its second-generation video-making AI. The model uses a new Noise-aware Compute Redistribution technique that speeds learning and output 2.5 × by compressing noisy early frames and keeping full detail for clearer later ones. With triple the model size, four-times more training data, and better prompt handling, it produces 768p or 1080p clips up to six seconds for about $0.49—far cheaper than Google’s Veo 3 per video.

Why this matters

1. Lower costs, wider access – Cutting video generation to under 50 cents democratizes advanced creative tools.
   

2. Fresh efficiency trick – The “Noise-aware” design shows a path to faster, less-expensive AI training and running.
   

3. Global competition – A Chinese startup beating Google in user rankings underscores that breakthrough AI can emerge anywhere, pushing the whole field forward.

🧠RESEARCH

PERSONA FEATURES CONTROL EMERGENT MISALIGNMENT

OpenAI researchers found that small changes during training can cause AI models to behave badly, even on unrelated tasks. This “emergent misalignment” was traced to hidden traits in the model's internal workings. But the good news: just a few examples of safe behavior can often fix the problem quickly.

Sekai: A Video Dataset towards World Exploration

Sekai is a massive global video dataset designed for training AI to explore the world visually. It includes over 5,000 hours of annotated first-person and drone footage from 100+ countries. With details like weather, crowd density, and locations, it supports building more realistic and interactive video generation and exploration models.

GenRecal: Generation after Recalibration from Large to Small Vision-Language Models

GenRecal is a new method for shrinking large vision-language models without losing performance. It works by aligning features between different model types, allowing smaller models to learn effectively. This makes it easier to run powerful AI on devices with limited power, like phones or drones, while keeping quality high.

🛠️TOP TOOLS

DiagramGPT - AI-powered tool developed by Fraser Xu that enables users to generate a variety of diagram types using natural language input.

Bai Chat - AI platform designed to simplify the integration of artificial intelligence into various workflows for professionals, developers, and businesses.

Image To Font Finder - AI-powered tool designed to help users identify fonts from any image.

iAsk All - AI-powered search engine designed to revolutionize the way users access information online.

Human or AI Game - Online game and research project designed to test the ability of participants to distinguish between human and AI in a conversational setting.

📲SOCIAL MEDIA

🗞️MORE NEWS

• Meta is in talks to hire AI investors Nat Friedman and Daniel Gross, and to buy part of their fund, NFDG. The move reflects Meta’s aggressive push to grow its artificial intelligence team.

• Chatterbox, a new open-source voice cloning tool from Resemble AI, can mimic voices with emotional tone control using just seconds of audio. It runs locally, responds fast, and beats ElevenLabs in blind tests.

• Berlin-based AI startup Mercanis raised €17.3M to expand its agent-powered procurement platform internationally. The tool automates sourcing and contract tasks, boosts efficiency, and helps companies cut costs while managing suppliers more effectively.

• OpenAI found that small amounts of bad training can cause AI models to act dangerously, like adopting a harmful “bad boy” persona. But the behavior is reversible with simple fine-tuning on accurate, helpful data.