Summer Ambitions in Menlo Park
In the summer of 2025, Menlo Park buzzes with ambition as a new lab for meta-superintelligence is taking shape. Mark Zuckerberg has laid out an audacious goal: to create an intelligence that surpasses human capabilities, or what he calls “superintelligence.” This new endeavor isn’t about crafting a distant oracle; it aims to deliver a “Personal Superintelligence” that everyone can benefit from. The ambitious multi-gigawatt AI supercomputer cluster is set to be named “Prometheus,” with plans for its debut in 2026.
The name Prometheus carries a profound mythological weight—a tale of enlightenment and hubris that revolves around the theft of fire from gods, leading to dire consequences. By naming their creation, Meta is, in a sense, drafting their own narrative, framing their achievements in heroic—or perhaps tragic—light. This narrative feels especially Californian; it embodies the ambition to create something transformative, albeit shadowed by the potential overreaching of natural boundaries.
These undertakings are envisioned as contemporary Apollo missions, focused not on the physical moon but on expanding human cognitive capabilities.
To chase this modern myth, executives have started to “turn things upside down.” Reports suggest that a crisis is brewing; four separate AI departments underwent major reorganization in just six months. An internal memo revealed that Meta was trailing in an “AI arms race” until this consolidation took effect. This restructuring was, by many accounts, viewed as urgent, with considerable budget implications. In fact, the capital expenditures for 2025 total approximately $72 billion, with Zuckerberg expressing intentions to invest even more. Notably, $27 billion in deals were finalized in October, including a joint venture with Blue Owl to develop a data center.
The competition for talent in 2025 has escalated to what feels like a frantic, almost desperate, race. By mid-August, Meta had recruited over 50 leading researchers from companies like OpenAI and Google, offering astonishing compensation packages that reportedly reach up to nine figures. Rumor has it that one AI lab co-founder turned down a staggering $1.5 billion offer from Meta, indicating just how intense things have become. This isn’t a tranquil recruitment drive; it’s a full-blown frenzy.
This flurry of activity aims to address existing shortcomings. The turning point arose not from euphoric discoveries, but from a general sense of dissatisfaction. During mid-2025, the release of the Llama 4 model family received mixed reviews, with the Behemoth model—a significant two trillion-parameter research initiative—seeing a lackluster response. To counteract this, Zuckerberg strategically redeployed personnel, disbanding a group that was not delivering results, and forming a more elite team. A new lab, intriguingly named TBD Lab, directed by the newly appointed Chief AI Officer, Alexandr Wang, has taken shape amidst these changes.
This TBD Lab stands as a central, protected hub. While many positions were cut in an effort to “reduce bureaucracy,” TBD Lab has been given full autonomy. The long-established FAIR research department, which once enjoyed a high degree of independence, is now positioned as the “innovation engine” supporting TBD. Additionally, the product team, led by Nat Friedman, has been tasked with connecting lab discoveries to market needs, while the infrastructure team is responsible for developing the massive computational backbone required.
The scale of this infrastructure is remarkable; we are not merely discussing servers in racks anymore. Reports indicate that this multi-gigawatt data center will cover “a significant portion of Manhattan.” It’s a quest reminiscent of the Apollo missions, but aiming for the human mind instead of the moon.
The evolving story from Meta is worth monitoring. Baite announces its vision for a “leaner, more efficient unit,” proclaiming itself a startup within the company. Interestingly, after an intensive hiring spree, Meta froze new recruitment not as a cost-saving measure—but to stabilize the new structure. It feels as though the turmoil surrounding massive expenditures, substantial salaries, and rejected offers might soon settle down, but who really knows?
Inside this streamlined organization, a cultural narrative is developing. There’s a clash between the new, high-priced talent poached from rivals and the longstanding Meta veterans who believed in the company’s earlier commitment to open-source science. This shift raises serious questions about the ethos that once set Meta apart. Zuckerberg hinted that the most potent models—those that might inch us closer to “superintelligence”—would remain proprietary, potentially for safety or strategic reasons. The transition from an open-source ideal to a closed, competitive model is evident.
Meanwhile, skepticism looms from the sidelines. Analysts express concerns that we may be experiencing an AI bubble, warning of “dilution of shareholder value without any clear innovation.” Investors are no longer dazzled by flashy showcases; they are seeking tangible returns. The ambitious vision of superintelligence clashes head-on with the immediate pressures of delivering durable, scalable products.
As companies and cultures begin to define their futures in such grand and mythical terms, the real question isn’t just whether we can create thinking machines—it’s about what those efforts reveal about ourselves: our ambitions and our vulnerabilities. Meta is at a pivotal juncture, propelled by substantial financial backing as abstract as the goals they’re pursuing, and possibly spurred by that most human emotion: the anxiety of being left behind.
