Inside Meta’s AI Revolt: How the Secretive TBD Lab and ‘Project Avocado’ Sparked a Civil War at Menlo Park

A profound cultural schism has erupted within the halls of Menlo Park, threatening to derail Mark Zuckerberg’s most ambitious pivot yet. As 2025 draws to a close, Meta Platforms is no longer just battling external rivals like OpenAI and Google; it is fighting a war against itself. At the center of this storm is the controversial “TBD Lab,” a secretive and elite division formed to chase superintelligence, whose rise has triggered high-profile departures, morale-crushing layoffs, and a fundamental shift in the company’s philosophy. The social media giant, once the champion of open-source artificial intelligence, is reportedly pivoting toward a closed, proprietary future with a delayed flagship model codenamed "Avocado," leaving its legacy workforce alienated and investors anxious.

The friction stems from a radical restructuring orchestrated by Zuckerberg earlier this year, marked by the stunning $14.3 billion acqui-hire of Alexandr Wang and his team from Scale AI. Wang, now Meta’s Chief AI Officer, was handed the keys to the newly formed TBD Lab, a unit explicitly designed to operate with the agility of a startup rather than the consensus-driven bureaucracy of a trillion-dollar corporation. This move effectively sidelined Meta’s long-standing AI division, Fundamental AI Research (FAIR), and its leaders. Reports confirm that the tension reached a breaking point this quarter, leading to the departure of AI pioneer Yann LeCun, a symbolic blow that has shaken the scientific community. LeCun’s exit, coupled with the slashing of 600 jobs from legacy AI and product teams in October, has drawn a clear battle line: you are either part of the new elite or you are obsolete.

Inside sources describe the atmosphere at Meta as a clash of civilizations. The TBD Lab operates under a grueling “70-hour workweek” culture, incentivized by eye-watering compensation packages that dwarf those of tenured staff. In return, Wang’s team has demanded and received priority access to computing resources, allegedly at the expense of established product divisions. This resource hoarding has sparked heated confrontations between Wang and veteran executives like Chief Technology Officer Andrew Bosworth and Chief Product Officer Chris Cox. While the old guard argues for integrating AI into existing platforms like Instagram and WhatsApp to drive immediate revenue, Wang’s faction is reportedly focused singular-mindedly on catching up to competitors with raw model performance, viewing current product constraints as a hindrance to achieving artificial general intelligence.

The strategic pivot is most visible in the development of "Avocado," Meta’s next-generation AI model. For years, Meta distinguished itself by releasing its "Llama" models as open-source software, a strategy that endeared it to developers worldwide. However, under the guidance of the TBD Lab, "Avocado" is being built as a closed, proprietary system, mirroring the guarded approaches of Google’s Gemini and OpenAI’s GPT series. This reversal has not only angered the open-source community but also confused internal teams who spent years evangelizing openness. Compounding the issue is the news that Avocado, initially slated for a late 2025 release, has been delayed to the first quarter of 2026 due to training hurdles, further fueling Wall Street’s skepticism about the return on Zuckerberg’s massive capital expenditures.

As Meta enters 2026, the stakes could not be higher. The company is effectively betting its future on a small, sequestered team of new hires while dismantling the culture that built its empire. If the TBD Lab succeeds, Zuckerberg will be vindicated as a ruthless visionary who successfully steered his company into the age of superintelligence. If it fails, or if the internal bleeding of talent continues, Meta risks losing its footing in the AI arms race entirely, left with a demoralized workforce and a delayed product that cannot compete with the market leaders. The "Year of Efficiency" has morphed into a year of upheaval, and the outcome of this internal civil war will likely define the next decade of the technology sector.