AWS
22 pieces on AWS, including 1 step-by-step guide.
Guides
News & Analysis
MCP 2026-07-28 Goes Stateless, Eliminating Sticky Sessions on AWS
MCP's July 28 2026 revision removes protocol-level sessions, eliminating ALB sticky routing and shared session stores for AWS deployments.
AWS AgentCore V2 Cuts Cold Starts to 2 Seconds Across All Image Sizes
AWS AgentCore runtime V2 delivers ~2s P75 cold starts from 200 MB to 2 GB images, versus up to 30s on V1, via snapshot-based initialization.
SageMaker HyperPod Inference Gateway Cuts TTFT P95 by Up to 98%
AWS's new EKS addon uses real-time KV cache and queue metrics to slash first-token latency by up to 98% over Kubernetes round-robin, with zero app changes.
Amazon Bedrock AgentCore Hosts MCP Apps With Interactive HTML Widgets
AWS publishes a reference architecture for MCP Apps on AgentCore runtime, delivering sandboxed HTML widgets inside ChatGPT and Claude without AWS credential exposure.
SageMaker Prefix-Aware Routing Cuts LLM P50 TTFT by Up to 77%
Amazon SageMaker's new PREFIX_AWARE routing lifts KV cache hit rates from ~25% to ~82% on long-context workloads, cutting P50 TTFT by up to 77%.
SageMaker Feature Store's UpdateRecord Ends Read-Modify-Write Cycles
AWS ships UpdateRecord for SageMaker Feature Store: atomic partial writes on up to 100 features, no GetRecord required, available in all regions today.
G7 Beats G5 and G6 on MoE Inference With Half the GPUs
AWS benchmarks show G7 (Blackwell) delivers 60.8% more throughput than G6 and $0.90/1M tokens on 30B MoE models with only two GPUs.
Amazon Textract Preprocessing Fixes RAG Failures on Complex Documents
AWS shows how interposing Amazon Textract before a Bedrock Knowledge Base eliminates hallucination and incomplete extraction on PDFs, spreadsheets, and images.
AWS Agent Registry Is Now Generally Available on Bedrock AgentCore
AWS Agent Registry hits GA, giving engineering teams a governed catalog for AI agents, tools, and skills with semantic search and EventBridge-wired approval workflows.
AWS Open Sources Kiro Crew: 39,000 Internal Users Before Public Release
AWS open-sourced Kiro Crew on Aug 30, 2026 — an async multi-agent coding framework used by 39,000 Amazon developers before external release.
Decathlon Cuts Forecast Error 15 pp With 120M-Param Chronos-2
Decathlon runs weekly demand forecasting across 15,000 SKUs in 75 seconds on a CPU instance for ~$0.03 per run, cutting WAPE by up to 15 percentage points.
SageMaker SDK v3 Replaces Dozen Estimator Classes With Two Primitives
AWS shipped SageMaker Python SDK v3 on 26 Aug 2026, collapsing framework-specific estimators into ModelTrainer and ModelBuilder with runtime code injection.
SageMaker HyperPod Gets Managed Ray on EKS, Replacing Manual Kubectl Setup
AWS launches managed Ray integration for SageMaker HyperPod on EKS, replacing manual YAML, port-forwarding, and Grafana setup with a console-driven workflow.
FBG Multi-Agent Support System Cuts Containment Gap 56%
Fanatics Betting and Gaming's AWS multi-agent system improved containment 56% and resolution 53% in two months, using Claude, Nova Lite, and custom RAG.
AgentCore Web Search Gains Per-Call Domain and Date Filters
AWS ships connector version 1.2.0 for AgentCore Web Search with per-call domain allowlists, denylists, and ISO-8601 date bounds — enforced server-side.
Amazon Bedrock AgentCore Payments Goes GA With MPP and Spending Caps
AgentCore Payments hits GA on August 18 2026, adding MPP protocol support, an x402 'upto' spending scheme, and infrastructure-layer transaction guardrails.
Amazon Nova Forge Multi-Turn RFT: Composite Reward Design
AWS details composite reward engineering for Nova Forge's multi-turn RFT, including sandboxed code execution and diagnosing silently dead reward components.
AWS Open-Sources Dogwood: Cedar Extended for Agent Tool-Call Sequences
Dogwood adds history-aware temporal conditions to Cedar, letting teams enforce sequence-level rules across agent tool calls. Apache 2.0, reference-only.
AWS Adds Native Vector Search to DynamoDB via SearchVectors API
DynamoDB now supports ANN queries via a new SearchVectors API, up to 4,096 dimensions, and per-byte billing — no separate vector store required.
Strands Robots + LeRobot + HF Buckets: One Record-Train-Deploy Loop
AWS and Hugging Face demonstrate a full robotics data loop: record episodes, sync with byte-level dedup, stream-train, and redeploy — all in LeRobot format.
OpenAI Daybreak Models Now Available on Amazon Bedrock
OpenAI's Daybreak Blue and Daybreak Red cybersecurity models are now accessible via Amazon Bedrock, requiring Daybreak Access enrollment.