tech
Building Abundant Intelligence
A full-stack approach to making advanced AI more capable, more affordable, and more widely useful.
TL;DR
- AI infrastructure's value lies in enabling more capable intelligence at a lower cost, benefiting all of humanity.
- A cycle of better intelligence, broader adoption, and increased investment drives progress and efficiency.
- Recent pricing reductions for GPT-5.6 Luna and Terra, along with speed improvements for GPT-5.6 Sol, expand customer flexibility.
- The focus is on the cost of a successful outcome, not just token prices, with stronger models potentially being more economical.
- Engineering work, including optimizing serving software and improving speculative decoding, reduces costs and increases efficiency.
- System improvements, such as better routing and context management, are crucial for AI efficiency, not just the models themselves.
- Real-world product use provides feedback that shapes research, strengthens products, and lowers serving costs.
- Massive user and business adoption deepens AI integration, transforming knowledge work and organizational operations.
- A coordinated system and cross-layer learning are key, with flexibility in owning, partnering, or buying components.
- Investment decisions are evidence-based, focusing on user growth, demand, and progress in model capability and efficiency.
- The ultimate goal is more useful intelligence within reach, characterized by increasing capability, affordability, and value.