// tag
#Ai
16 posts
Agent Authorization Needs to Understand Time and Sequence
Amazon Bedrock AgentCore temporal policies extend authorization beyond identity so agent actions can be checked against workflow history, freshness, approvals, and cumulative limits.
Beyond the SDK Demo: Building Production-Ready .NET Agents with AgentCore
Moving a .NET AI agent from an SDK sample to Amazon Bedrock AgentCore means making deliberate choices about streaming, sessions, testing, deployment, and portability.
Persistent Compute Changes the Bedrock AgentCore Boundary
Amazon Bedrock AgentCore runtime instances bring persistent sessions and shared compute to agent workloads, changing the trade-off between stateless scale and durable execution.
Response Streaming for .NET on Lambda Changes Perceived Latency
Response streaming for .NET on AWS Lambda can deliver tokens and large responses earlier, but buffering, headers, timeouts, and disconnects become part of the API contract.
Agent UIs need protocols, not only chat boxes
AG-UI with Amazon Bedrock AgentCore shows how agent frontends can support generative UI, shared state, and human-in-the-loop workflows without coupling every frontend to one agent framework.
Autonomous incident resolution needs boundaries before autonomy
AWS DevOps Agent with Datadog MCP Server points toward AI-assisted incident response, but production teams need permissions, approvals, rollback, and observability boundaries before autonomous fixes.
AI-powered resilience testing is useful when it discovers real dependencies
An AWS resilience framework using Resilience Hub, Fault Injection Service, Systems Manager, and Bedrock AgentCore shows how teams can move from assumed reliability to continuous validation.
AI-driven migration orchestration is useful when it reduces handoff friction
AWS guidance on AI-driven large-scale server migration shows how Transform, Cloud Migration Factory, MGN, AgentCore, and Kiro can reduce coordination overhead across migration waves.
AWS Continuum points to security workflows that reason before they remediate
AWS Continuum for code vulnerabilities previews a security model that prioritizes, validates, and recommends remediation with context before moving toward automation.
Bedrock resilience patterns make LLM availability an architecture concern
AWS guidance on Amazon Bedrock and LLM gateway resilience shows why production generative AI systems need quota isolation, cross-Region routing, and multi-model fallback design.
Claude Sonnet 5 on Bedrock should trigger model selection reviews
Claude Sonnet 5 availability on Amazon Bedrock gives builders another production model option, but adoption should be driven by evaluation, cost, latency, and agent reliability.
EC2 G7 instances are a reminder to size GPU workloads by bottleneck
Amazon EC2 G7 instances bring NVIDIA RTX PRO 4500 Blackwell GPUs to AWS, but builders should evaluate them by memory, networking, storage, and software fit instead of GPU generation alone.
Bedrock managed entitlements make model access a platform control
Amazon Bedrock managed entitlements let organizations subscribe to marketplace models centrally and distribute access across accounts without broad AWS Marketplace permissions.
Lambda durable functions fit the messy middle of agent workflows
AWS Lambda durable functions give multi-agent and human-in-the-loop workflows checkpointing, replay, callbacks, and polling without forcing every team to assemble custom orchestration infrastructure.
Running pgvector on Aurora is a production operations decision
AWS guidance on pgvector in Amazon Aurora PostgreSQL highlights that vector search is not only a model feature; it needs indexing, memory, partitioning, and observability discipline.
Lambda MicroVMs make isolated sandboxes a serverless design choice
AWS Lambda MicroVMs give builders a new option for running user-generated and AI-generated code with VM-level isolation, fast resume, and controlled lifecycle state.