Frontier models aren’t simply producing better outputs. Advances in reasoning, coding, tool use and long-context understanding are changing which AI use cases are viable and how developers need to design the systems around them. Peter from Anthropic will unpack what these advances mean in practice: where models can now handle more complex work, where their limitations still surface, and the architectural decisions teams must make around reliability, evaluation, scale and control. Leave with a clearer understanding of which capabilities are ready to build on, the trade-offs that come with them and how to turn progress at the model layer into AI applications that deliver real-world value.