AWS Weekly Roundup: AWS Lambda Web Functions, Amazon S3 on the Vercel Marketplace, and more (October 19, 2026)
~2 min read
devopsawsweekly-rounduplambdas3aitrainium
Provision S3 general purpose buckets, S3 Tables, and S3 Vectors from the Vercel dashboard or by describing what you want in v0. It joins existing integrations for Aurora, DynamoDB, and OpenSearch Serverless — the whole stack in one AWS account with one bill. The good part: a short-lived OIDC token is exchanged at runtime with AWS STS for credentials scoped to that resource, so no static keys are stored anywhere.
The newest burstable generation reached South America — the unglamorous half of a workload: login gateways, small databases, CI/CD runners, low-traffic sites. T8i delivers up to 30% better price performance than T3, with up to 70% higher compute and 2.4x the EBS bandwidth. Same CPU credit system as T3, so migration is a size change rather than a redesign; t8i.micro and t8i.small are under the Free Tier.
Deploy Beam, Reflection AI's open-weight model for agentic coding and long-horizon tasks, from SageMaker JumpStart. A sparse MoE with 501B total / 23B active parameters, Apache 2.0 on Hugging Face, ~1 million-token effective context. Start with a JumpStart real-time endpoint or SageMaker HyperPod Inference with SGLang/vLLM for Kubernetes-native control. Also ported and optimized for AWS Trainium3, launching in N. Virginia and Ohio.
Write multi-step Go workflows that checkpoint their own progress and pause for up to a year waiting on an external event — deployment rollouts with approval gates, pipelines that call a model and then stop for human review. Ordinary Go tooling: go get, a single static binary on provided.al2023, and local testing with the durabletest package, no network or credentials required.
Define routes in CloudWatch to send logs to the destinations you choose, with chained parsers to transform them along the way.
The durable-execution SDK for Go rounds out Lambda's story for stateful workflow orchestration, and Beam on Trainium3 is worth watching if you want open-weight agentic coding models on self-managed inference infrastructure.