The High-Stakes Head Hunt Begins
Mark Zuckerberg is back in “founder-mode.” Over the past month the Meta CEO has fired off a flurry of WhatsApp DMs and surprise video calls to the world’s top AI scientists, asking for “just fifteen minutes” of their time. His pitch is blunt: Come build super-intelligence with me, sit next to my desk, and I’ll bankroll every risky idea you’ve ever had.
Researchers at Google, DeepMind, Anthropic even a few at OpenAI confirm they’ve seen some version of that message. Many decided to listen. Some have already jumped ship.
Why Zuckerberg Hit the Panic Button
Internally, Meta’s once-proud Llama-4 model is lagging. A “Behemoth” follow-up has slipped into 2026, and engineers quietly use Anthropic’s Claude for coding tasks. Zuckerberg, frustrated by the drag, scrapped the old roadmap and decided to start over with himself in the driver’s seat. Bloomberg’s reporting (echoed by The Decoder) says a WhatsApp chat titled “Recruiting Party” now pings around the clock, listing dream hires and who’s texting them next.
Follow the Money: Nine Figure Offers & a $14.3 B Bet
Poaching elite brains isn’t cheap. Meta is dangling pay packages “well into the eight-figure range,” and in rare cases, nine figures. Think NBA-superstar money for PhDs. The most dramatic move so far: a $14.3 billion splash for data-labeling powerhouse Scale AI, a deal sweet enough to pull founder Alexandr Wang out of his own company. One Meta insider joked the price was “fourteen Instagrams.”
The Daily Beast adds that roughly 50 seats have been reserved for this new “Superintelligence Group,” and every chair now sits within eyesight of Zuckerberg’s glass office.
Meet the “Fantastic 50”
Confirmed recruits already read like a fantasy league roster. Wang is tipped for Chief AI Officer. DeepMind veteran Jack Rae will steer pre-training. Several researchers from Google Brain and FAIR followed, lured by Meta’s promise of “Tesla-scale” GPU clusters and the autonomy to publish. Interview venues feel surreal: Lake Tahoe, Palo Alto backyards, a private hangar near Menlo Park. Candidates describe a CEO who “takes notes himself and knows your entire citation list.” If accepted, they sign on the spot.
Superintelligence, Data Centers & the Shelved “Behemoth”
What will this new group actually build? Two words keep popping up: “agentic reasoning.” The charter is to leapfrog today’s chatbots and chase AGI head-on. Meta has earmarked “tens of billions” in 2025 cap-ex for fresh compute, plus plans for multi-gigawatt data centers running custom accelerators. Longer term, Zuckerberg told recruits he’s prepared to spend “hundreds of billions” if that’s what it takes.
The shelved Behemoth model may get revived or replaced entirely once the team sets its own architecture.
How Rivals Are Reacting and Freaking Out
The shockwaves are immediate. On the eve of Meta’s offer window, Sam Altman published a blog post reminding staff that OpenAI is “before anything else, a super-intelligence company.” Google fast-tracked a new Senior VP title for DeepMind CTO Koray Kavukcuoglu to keep him from defecting. Internally, Amazon updated its “critical talent” retention budget. One DeepMind researcher joked, “Zuck just set off a salary nuke.”
Risk Factors: Ego, Culture Clashes & Regulation
Big personalities plus blank checks can implode. Meta’s legacy AI teams wonder where they fit, and veteran scientist Yann LeCun a vocal skeptic of current generative tech sits outside the new org chart. Regulators will also scrutinize a $14 billion bolt-on and nine-figure paydays during an election cycle already wary of AI’s influence. Even infrastructure poses risks: energy-hungry data centers could draw environmental fire, especially in California. Still, Zuckerberg seems willing to embrace the turbulence if it buys him a moon-shot.
What Happens Next and Why It Matters
Expect an official reveal once a “critical mass” of signatures lands likely before summer ends. Early prototypes may surface in Meta’s internal hackathons by Q4, with a public demo rumored for F8 2026. Whether the plan succeeds or stalls, the playbook is clear: own the best brains, feed them infinite compute, and swing for AGI. If Zuckerberg pulls it off, Meta could vault from LLM laggard to the company shipping the world’s first truly agentic assistant. If he fails, the effort might still reset salary norms, talent flows, and regulatory debates for everyone else.
Either way, the AI arms race just found its next gear and you can hear the whine of those GPUs from here.
Comments 1