Meta Platforms

Meta is the most-watched 2026 case study in what happens when an incumbent tries to buy itself an AI lab from outside its own talent pool. The Scale AI acqui-hire installed alexandr-wang as Chief AI Officer running Superintelligence Labs (SSL); within four months retention had fallen to 64%, eight named researchers had left, and andrew-bosworth had stood up a parallel Applied AI Engineering org under VP Maher Saba that reports around Wang rather than to him.

The May 2026 layoff wave (8,000 company-wide, ~600 inside SSL specifically) is the visible moment, covered in meta-ai-employees-miserable. The pattern around it — Yann LeCun publicly calling Wang "inexperienced," Zuckerberg's $135B AI capex commitment, the Fortune report on screen-and-keystroke tracking to train internal tools — is the more important story.

The earlier Meta is the one that made open weights a corporate strategy. Zuckerberg's July 2024 letter released with Llama 3.1, open-source-ai-is-the-path-forward, is still the statement people cite for why a large company gives weights away; Nathan Lambert's open-source-ai-reading-list calls it one of the clearest. By 2026 the bet had unwound. Bill Gurley reports Llama 4 Behemoth shelved and a closed Muse Spark in its place (from-open-source-software-to-open-source-strategy), and the ATOM Report counts Meta's share of new fine-tunes falling from a 44% peak to 11% as Qwen took over (atom-report-open-model-ecosystem).

Useful for cross-referencing whenever Meta-specific decisions show up: AI personnel moves, Llama model releases, Reality Labs re-org news, internal-mandate stories.

Cross-references: ai-bubble-pale-horses, ceo-ai-psychosis, ai-sycophancy-loop, tokenmaxxing.