PLATFORMS Signal 491
AWS Weekly Roundup: AWS Heroes Summit, Web Search on Amazon Bedrock, Dogwood, Kiro Crew, and more (August 10, 2026)
↗
AWS announced several platform updates including web search for OpenAI models on Bedrock, vector search for DynamoDB, higher Lambda network bandwidth, an open-sourced AI agent governance language called Dogwood, and a multi-agent collaborative workspace called Kiro Crew.
The additions to Bedrock and DynamoDB reduce the infrastructure you need to wire up for production AI agents, while the Lambda bandwidth increase removes a practical bottleneck for data-intensive serverless workloads. The open-sourcing of Dogwood and support for Agent Plugins signal AWS pushing toward portable, governable agent ecosystems rather than purely proprietary tooling.
Written by elseif from the cluster below · every claim links back to a sourceThe three things worth knowing
Amazon Bedrock now lets OpenAI models browse the web for real-time information while keeping data residency inside your AWS environment with zero data egress.
DynamoDB gains native vector search, eliminating the need for a separate vector database when building semantic retrieval into AI agent memory.
AWS Lambda functions outside a VPC with 2 GB or more memory now scale network bandwidth up to 3,000 Mbps at 10 GB, up from a previous ceiling.
THE READ
What the cluster adds up to.
The Bedrock web search capability is the most immediately consequential launch for teams building AI applications. OpenAI models available through Bedrock can now retrieve live internet content, which means agents can answer questions grounded in current information rather than relying solely on training data. The zero-data-eggress claim matters for regulated workloads where keeping traffic inside AWS boundaries is a compliance requirement rather than a preference.
Vector search landing directly in DynamoDB changes the architecture decision for teams that were previously forced to run a separate vector store alongside their primary database. You can now store and query embeddings alongside existing DynamoDB data with predictable performance, which simplifies the operational surface for agentic grounding workflows. The cost tradeoff is not detailed in the announcement, but the elimination of a separate managed vector database is a concrete reduction in moving parts.
The Lambda bandwidth increase from a prior ceiling to 3,000 Mbps at 10 GB memory addresses a known limitation for data-intensive serverless workloads. Functions outside a VPC configured with 2 GB or more memory now get proportional scaling from 625 Mbps upward. This matters for workloads that move large payloads between Lambda and other AWS services, though functions inside a VPC are not mentioned and may still face the previous constraints.
Dogwood, the open-sourced governance language for AI agents, introduces temporal policy conditions to Cedar, meaning authorization decisions can factor in the history of an agent's actions within a session rather than evaluating each request in isolation. This is a meaningful shift for teams concerned about agent safety, as it enables policies that detect escalation patterns across a conversation. Agent Plugins, a separate but related announcement, provides a vendor-neutral packaging format for agent extensions that work across Kiro, VS Code, and Cursor, reducing lock-in to any single IDE or toolchain.
Kiro Crew introduces a persistent, multi-agent workspace within the Kiro IDE that spans repos, tools, and days rather than being scoped to a single chat session. It supports parallel agent efforts and delegation to subagents that report back, which positions it as infrastructure for longer-running engineering workflows. The AWS Transform continuous modernization reaching general availability rounds out the platform news, offering automated, ongoing remediation of technical debt across source repositories rather than treating modernization as a one-time migration event.
Written by elseif from the cluster below · checked for specifics the sources never containedTHE CLUSTER