Check your email (including spam) for our email with the subject: “Complimentary subscription to Yaro on AI and Tech Trends | Your Top AI Newsletter.” Please reply to it with a quick “Hello” to include you in our Substack newsletter as we move towards that platform. Thanks.

Prediction markets are taking over our lives, as anyone can basically bet even as little as a $1 on any future event. As these platforms generate billions in trading volume, Wall Street is beginning to pay attention and explore similar financial products. At the same time, “world models” are becoming one of the hottest areas in AI, with startups raising over $2B in recent weeks to build systems that help machines understand and simulate the physical world. We break down five major approaches shaping this emerging field.

Meanwhile in Washington, Sen. Elizabeth Warren is questioning how safe classified government data is when shared with AI companies, raising concerns about security risks as the Pentagon deepens partnerships with frontier AI developers.

Let’s get into it. Stay curious.

• 🧰 AI Tools - Top 2026 AI Tools as per Zapier

• The Race to Build “World Models” Is Splitting Into Five Approaches

• 📚Learning Corner - All OpenAI Live courses and webinars.

• Wall Street Moves Into Prediction Markets

• Warren Questions Pentagon Deal With xAI

Get 60% off for 1 year

📰 AI News and Trends

• NVIDIA GTC 2026 kicks off today with Jensen Huang's keynote, 39,000 attendees, and $600B AI infrastructure spend in focus.

• Morgan Stanley warns a transformative AI breakthrough is imminent in H1 2026, with GPT-5.4 already scoring at a human-expert level.

• OpenClaw (acquired by OpenAI) becomes GitHub’s #1 most-starred project, surpassing React & Linux, agentic AI goes mainstream

• Tech companies are spending an unprecedented $700 billion this year on AI data centers. Jensen Huang: “We’re not anywhere near the peak.”

• ByteDance pauses global launch of Seedance 2.0 AI video generator.

• Pet dad turns to ChatGPT to create a cancer vaccine for his dying dog.

• OpenAI’s Bid to Allow X-Rated Talk Is Freaking Out Its Own Advisers.

Other Tech News

• Meta is reportedly considering layoffs affecting up to 20% of its workforce.

• The Trump administration is set to receive roughly $10 billion from investors in the recently completed deal to take control of TikTok’s U.S. business, according to the Wall Street Journal.

• Anduril wins up to $20B, 10-year US Army contract.

• Facebook makes it easier for creators to report impersonators.

Share

The Race to Build “World Models” Is Splitting Into Five Approaches

AI startups are pouring huge capital into “world models” systems that help AI understand and simulate reality. In the past few weeks alone, AMI Labs raised $1.03B, and World Labs raised $1B, but the term “world model” now covers several very different technical strategies. Zhuokai Zhao shared on X

1. JEPA latent prediction
Led by Yann LeCun and AMI Labs. Instead of predicting pixels, JEPA models predict abstract representations of the world, making them far more efficient. One model trained on 1M+ hours of video and needed just 62 hours of robot data to perform zero-shot planning.

2. Spatial intelligence (3D worlds)
Fei-Fei Li is betting that AI needs explicit 3D understanding of environments. World Labs’ system generates persistent 3D scenes that can be explored and edited like real environments.

3. Learned simulation
Companies like Google DeepMind are building models that simulate worlds directly. Systems such as Genie and Dreamer allow AI agents to learn by interacting with simulated environments, even achieving complex goals like finding diamonds in Minecraft without human data.

4. Physical AI infrastructure
NVIDIA is building the platform layer. Its Cosmos stack trains world models on 20M hours of video (9,000T tokens) and provides tools for robotics, autonomous systems, and physical AI.

5. Active inference
VERSES AI uses a neuroscience-inspired approach where agents minimize “surprise” in the world using probabilistic reasoning instead of deep learning.
World models may become the next major AI architecture after large language models, powering robotics, simulation, and autonomous systems, but the industry is still experimenting with multiple competing approaches.

Share

📚Learning Corner

• All OpenAI Live courses and webinars.

Wall Street Moves Into Prediction Markets

Prediction markets, platforms where traders bet on future events, are gaining traction in traditional finance.

Trading platform Tradeweb recently partnered with Kalshi and saw overwhelming interest from institutional clients. Kalshi says billions in trading volume already come from investors trading contracts tied to climate, tech, and macro events.

Major finance players are entering the space:

• Intercontinental Exchange invested $2B in Polymarket

• Jump Trading and Susquehanna International Group are providing liquidity

• Nasdaq is exploring event-based trading contracts

Prediction markets are being used to forecast elections, interest rates, wars, and economic indicators, turning collective market bets into real-time forecasting tools.

Investor takeaway: If regulators approve new products like prediction-market ETFs, the sector could evolve into a new derivatives asset class on Wall Street.

Share Yaro on AI and Tech Trends | Your Top AI Newsletter

🧰 AI Tools of The Day

• The best AI productivity tools in 2026, as per Zapier.

Warren Questions Pentagon Deal With xAI

Sen. Elizabeth Warren is demanding answers from Defense Secretary Pete Hegseth about reports that the Pentagon may allow xAI to operate on classified military networks.

Warren warned that xAI’s chatbot Grok could pose cybersecurity and safety risks, including leaking classified information, producing inaccurate responses, or lacking proper safeguards for military use. She requested documentation showing how the Defense Department evaluated Grok’s security and data-handling practices.

The concern comes after the Pentagon awarded xAI a contract worth up to $200 million to develop AI systems and reportedly reached a deal in February to expand its use on secure networks. Grok has faced scrutiny from regulators and governments worldwide over weak guardrails and controversial outputs, including manipulated images and antisemitic responses.
Doge may have already given xAI and all its vertical companies unprecedented access to sensitive government and citizen data. And now that Anthropic shows us that AI companies can be ethical, the governments are rushing to find less ethical companies to replace Claude AI to adopt frontier AI. Policymakers are increasingly questioning whether emerging models are secure and reliable enough for national security systems.