.png){kind=logo}
Apple's AI journey just hit a significant speed bump. Four key engineers from its Foundation Models team have left for Meta's Superintelligence Lab. The exits include Ruoming Pang, who built the Foundation Models group, and key early contributors Tom Gunter, Mark Lee, and Bowen Zhang.
These engineers set essential technical directions about the internal architecture of Apple AI, particularly for Siri and Apple Intelligence features. Their sudden departure has now strained the company's vision of in-house AI just as it is preparing to launch the iPhone 17 series. Internally, this movement has raised eyebrows, especially since some team members reportedly requested more transparency and faster progress but received silence.
Insider sources state that engineers were offered multi-million-dollar compensation packages, including equity, bonuses, and roles on high-impact AI projects. Meta's clear roadmap toward creating scalable, multimodal AI systems with real-world applications makes it even more appealing.
At Apple, the AI rollout has been slower. Despite big promises at last year's WWDC, many Apple Intelligence features are still pending or available only in beta. Siri, in particular, struggles with natural language understanding, something the company hoped its Foundation Models team would fix.
However, with internal delays and a lack of significant talent, the tech titan is now putting its weight behind partnerships with OpenAI and Anthropic. This might solve temporary issues but highlights a more permanent problem: Apple's in-house models may never be ready to compete with GPT-4 or Meta's LLaMA.
The company's strengths have always been refinement, design, security, and seamless integration. But Apple AI development moves differently. It thrives on openness, iteration, and talent density. Losing core engineers from a foundational team isn't just a hiccup; it signals deeper friction.
In contrast, Meta is building momentum by injecting funds into infrastructure and emphasizing speed over secrecy. Meta appears like a creative playground to engineers who want to do extraordinary work.
The tech leader has a long history of turning second-mover status into a long-term win. This laptop innovation and the M1 chip came after years of work and far exceeded expectations. The question now is whether the tech giant can do the same with AI.
To achieve this, it must develop quickly by giving its teams freedom, rewarding innovation, and building AI from the inside. It's not over yet, but Apple needs to get moving.
