We believe the next wave of consumer experiences will be vastly different to those of today. As consumer hardware continues to evolve, software will exist across multiple surfaces, where visual, text and audio are interoperable. Users will be able to interact with products using any modality they choose, whenever they want.

To make this future possible, the cost structure of AI has to fundamentally change. Every generational consumer product has taken a previously impossible creative capability and made it free for everyone.

This isn't possible today for AI-native experiences, because AI inference is many orders of magnitude more expensive than traditional software. Therefore, to make consumer AI free at scale, marginal inference costs must be driven to zero.

Duration exists to build towards this, with two areas of focus:

Applied AI Research: exploring how multi-modal experiences can be delivered at zero incremental cost through on-device inference.

Product: building the consumer products on that research, creating experiences that couldn't exist before.

This structure creates a flywheel: data and evaluations from our products accelerate our research, and our in-house models and breakthroughs let us build products no one else can.