Meta’s engineers spent the first 20 years under a “move fast and break things” ethos, later refined to “move fast with stable infra.” Back in the 2010s, Facebook even printed those mantras in a 70-page “little red book” and plastered them across campus: “Done Is Better Than Perfect,” “Fail Harder,” “Fortune Favors the Bold.” By 2022, loose processes, minimal testing, and an engineer-driven CEO made Meta one of Big Tech’s most product-centric outfits, where small teams rolled out features in days and auto-rollout systems scaled Threads to 100 million users in its first week. Then AI fever hit. Mark Zuckerberg poured billions into Llama models—Llama 1 in February 2023, Llama 2 in June 2023, Llama 3 spring 2024 and the disappointing Llama 4 in April 2025. Meta shelled out $14.8 billion for a 49% stake in Scale AI, recruited its CEO Alexandr Wang to overhaul AI strategy, and tried (unsuccessfully so far) to acquire China’s Manus AI for $2 billion. Engineers now face constant pressure to adopt these in-house tools, often in half-baked form, while top leadership chases every emerging platform trend after VR and the metaverse underperformed. That shift turned software teams from profit generators into cost centers almost overnight. Reports describe mass reorganizations, public embarrassment over a recent outage, and “AI psychosis” in which executives demand AI fixes for problems that don’t exist. Core engineers feel sidelined, processes contradict each other, and morale has plummeted. What once looked like a high-performance machine now resembles an internal mess driven by misplaced priorities.