Sunil Pai and Matt Carey describe how persistent, addressable objects can support long-running AI agents, scheduled tasks and connections that resume across clientsDurable workflow execution preserves progress so a multi-step process can resume after interruptions using recorded completion state.. They use a shared state and streaming example to explain the difference from a stateless request handler.
The speakers explain dynamic workers as isolated environments for running generated codeAn AI agent sandbox is an isolated execution environment that limits which files, processes, networks, credentials, and external systems an agent can access.. They describe limiting the APIs and network accessAn agent permission boundary defines and enforces which resources and actions an automated agent is allowed to use. exposed to that code, and discuss generated interfaces as one possible application.
Sunil Pai and Matt Carey also discuss code mode, Model Context Protocol serversModel Context Protocol is a standard way for AI applications to connect to external tools and sources of context., resumable streams and possible future agent features. Their performance and security assessments are presented as the speakers' claims.
Watch on YouTube
