AWS hosted its annual invite-only Heroes Summit last week, bringing together global experts in AI, serverless, and containers for direct collaboration with AWS product teams. The event featured a fireside chat with CEO Matt Garman, an AMA with James Hamilton, and breakout sessions focused on feedback and knowledge sharing.

Several launches caught attention. Amazon Bedrock now supports web search for OpenAI models (GPT-5.4, GPT-5.5, GPT-5.6), allowing AI applications to pull real-time internet data while keeping data residency within AWS and no egress. AgentCore now offers dedicated runtime instances for AI agents, giving more control over execution environments and predictable cost. DynamoDB added vector search, letting users store and query embeddings alongside existing data without a separate vector database, supporting semantic retrieval for agent memory.

AWS Transform continuous modernization is now generally available, helping teams analyze and remediate technical debt across codebases at scale, including mainframe and legacy workloads. AWS Lambda now supports up to 3,000 Mbps network bandwidth for functions with 10 GB memory, scaling from 625 Mbps at 2 GB, enabling faster data-intensive workloads.

On the open-source front, AWS introduced Dogwood, a governance language for AI agents that adds temporal conditions to Cedar policies, and AgentCore now supports temporal policies based on an agent's action history. AWS also backed Agent Plugins, an open standard for portable agent extensions that can be packaged once and used across tools like Kiro, VS Code, and Cursor. Kiro Crew, a persistent workspace in the Kiro IDE, enables collaborative multi-agent development across repos and days, with support for parallel efforts and subagents.

For full details, AWS points to its What's New and AWS Blogs pages, plus upcoming events and the AWS Builder Center.

Source: AWS