Zuckerberg's $15 billion push for AGI

PLUS: OpenAI turns to a rival, an AI math genius, and NVIDIA forecasts the climate

Good morning, AI enthusiast.

Meta is making a massive $15 billion move to acquire a nearly 50% stake in data-labeling giant Scale AI. This investment is the cornerstone of Mark Zuckerberg's new, personally-led superintelligence group aimed at accelerating the race to AGI.

The deal brings Scale AI’s CEO into Meta's fold to help lead the initiative, signaling a major talent and capability acquisition. With Meta’s own flagship model reportedly delayed, is this strategic maneuver the key to leapfrogging its rivals?

In today’s AI recap:

  • Our new AI training video (watch here)

  • Meta's $15B AGI bet on Scale AI

  • OpenAI taps rival Google for cloud compute

  • An AI model stuns top mathematicians

  • NVIDIA’s new AI climate forecaster

Zuckerberg's $15B AGI Bet

The Recap: Meta is reportedly investing nearly $15B for a 49% stake in data-labeling giant Scale AI. This massive deal is part of an aggressive push to build a new "superintelligence" group and accelerate its race toward artificial general intelligence (AGI).

Unpacked:

  • Mark Zuckerberg is personally assembling a new ~50-person lab dedicated to AGI, driven by frustration with Meta's current pace in the AI race.

  • The deal reportedly brings Scale AI's CEO, Alexandr Wang, into Meta to help lead the new superintelligence initiative, signaling a major talent acquisition.

  • This move comes as Meta feels pressure from rivals, especially after delaying its own flagship "Behemoth" model amid concerns about its performance.

Bottom line: This is more than an investment; it is a strategic maneuver to secure both Scale AI's essential data-labeling capabilities and top-tier leadership. By acquiring a massive stake instead of attempting an outright purchase, Meta accelerates its AGI ambitions while likely aiming to navigate complex regulatory hurdles.

OpenAI Taps Google Cloud in a Surprising Twist

The Recap: In a surprise move, OpenAI is now using Google's cloud infrastructure, diversifying its resources beyond its primary backer, Microsoft. This partnership between fierce rivals highlights the colossal compute demands of modern AI development.

Unpacked:

  • The deal addresses OpenAI's surging demand for computing power, a need underscored by its revenue which recently surged to $10 billion on an annualized basis.

  • For Google, landing OpenAI as a customer is a major win that validates its strategy of acting as a neutral provider of high-performance compute, powered by its in-house Tensor Processing Units (TPUs).

  • This is the latest step in OpenAI's plan to diversify beyond Microsoft, whose exclusive cloud provider status ended in January, and follows other infrastructure deals with partners like Oracle and CoreWeave.

Bottom line: The astronomical need for AI computing power is forcing even the biggest competitors into pragmatic alliances. This signals that the underlying infrastructure layer is a critical battleground, with cloud providers racing to power the entire AI ecosystem.

AI Training

The Recap: In this video, we walk through how we used n8n + AI to fully automate a daily AI-focused newsletter that gets read by tens of thousands of people. From scraping hundreds of news articles across the internet every day, to generating bite-size content with custom prompting, this video shows exactly how you can automate an industry-specific newsletter for your niche!

P.S We also launched a free community for AI Builders looking to master the art and science of building AI Automations — Come join us!

AI Stuns Top Mathematicians

The Recap: In a private meeting, OpenAI's latest reasoning model left 30 of the world's leading mathematicians in awe by solving problems they designed to be unsolvable by AI, using the new FrontierMath benchmark.

Unpacked:

  • The test was created by the nonprofit Epoch AI, which was commissioned with funding from OpenAI to develop hundreds of novel, high-level math problems that had not been published online.

  • The model, o4-mini, is a lighter and more nimble reasoning model trained on specialized data to make intricate deductions, moving beyond the simple text prediction of older LLMs.

  • In one instance, a mathematician watched the model tackle a Ph.D.-level problem by first studying the relevant literature, solving a simpler version to learn, and then presenting the correct final answer in just ten minutes.

Bottom line: This marks a significant step for AI, moving it from a generalist tool to a potential collaborator in deeply specialized scientific fields. The future role of human experts may shift from finding solutions to defining the most important questions for AI to solve.

Where AI Experts Share Their Best Work

Join our Free AI Automation Community

Join our FREE community AI Automation Mastery — where entrepreneurs, AI builders, and AI agency owners share templates, solve problems together, and learn from each other's wins (and mistakes).

What makes our community different:

  • Real peer support from people building actual AI businesses

  • Complete access to download our automation library of battle-tested n8n templates

  • Collaborate and problem-solve with AI experts when you get stuck

Dive into our course materials, collaborate with experienced builders, and turn automation challenges into shared wins. Join here (completely free).

NVIDIA's AI Climate Forecaster

The Recap: NVIDIA has unveiled cBottle, a generative AI model that simulates the global climate at an unprecedented kilometer-scale resolution. It promises to make climate predictions thousands of times faster than traditional methods.

Unpacked:

  • The model runs on NVIDIA’s Earth-2 platform and is thousands of times faster and more energy-efficient than older numerical models, reducing petabytes of data by up to 3,000x for a single sample.

  • It can perform tasks like correcting biased models, super-resolving low-resolution climate data, and filling in missing or corrupted information based on learned patterns.

  • Researchers can now experiment with cBottle directly through its openly available codebase on GitHub.

Bottom line: This marks a significant leap in our ability to predict complex climate patterns with greater speed and detail. The development of high-fidelity digital twins of Earth will enable better preparation for extreme weather and long-term climate change.

The Shortlist

Mistral announced Magistral, its first family of reasoning models, releasing the 24B parameter Magistral Small as an open-source model alongside a more powerful enterprise version.

BitBoard launched out of Y Combinator with AI agents designed to automate repetitive back-office tasks for healthcare clinics, such as prepping charts and managing patient referrals.

Microsoft made OpenAI's Sora text-to-video model available for free to all users through its new Bing Video Creator feature on its iOS and Android mobile apps.

Reports detail how Google's AI Overviews feature is significantly reducing referral traffic to news publishers, with some outlets seeing search-driven traffic decline by over 50%.

What did you think of today's email?

Before you go we’d love to know what you thought of today's newsletter. We read every single message to help improve The Recap experience.

Login or Subscribe to participate in polls.

Signing off,

David, Lucas, Mitchell — The Recap editorial team