Meta AI reorg turmoil and repeated Muse Spark developer release delays
Meta's $META AI overhaul is stumbling in public. CTO Andrew Bosworth told staff the company did an atrocious job on its AI reorg, per an internal memo seen by Wired. Its new Muse Spark model, built to close the gap with rivals, keeps slipping its developer release over bugs.
Meta $META is struggling to project confidence in its AI strategy at the exact moment it is spending heavily to catch up. In an internal memo seen by Wired, Chief Technology Officer Andrew Bosworth acknowledged that the company did an atrocious job explaining its vision and supporting employees through a sweeping AI reorganization. That restructuring created a new Applied AI division, and the upheaval has unsettled staff who are uncertain about their roles and impact. The internal friction lands alongside a product problem. Muse Spark, unveiled in April as the first model from Meta's Superintelligence Labs and positioned to close the gap with leading rivals, has had its developer release delayed repeatedly. Reports point to a mix of software bugs and underlying infrastructure issues, with no firm launch date set. Until outside developers can test the model, there is no independent read on whether Meta has regained ground in frontier AI, a position it lost when Llama 4 underwhelmed. The combination matters because Meta is committing enormous sums to AI talent and data center capacity without a cloud business to recoup those costs directly. Its main payoff route runs through advertising. For now, the public picture is one of internal disorder and a flagship model that is not yet in developers' hands.