Off-by-none: Issue #351

February 3, 2026

And we're back! 🎉

In our previous issue, we rang in the new year, speculated on the future of AI and serverless, and shared plenty of posts from the community. This week, AWS de-stresses developers with account names in the console, DynamoDB lets you cross (account) streams, and Databricks puts a new spin on serverless databases. Plus, we have a month's worth of amazing community posts to catch up on.

News & Announcements

AWS shipped a lot this past month, with a noticeable focus on quality-of-life improvements and more AI agent-ready building blocks. Out of all the announcements, I think my favorite is that the AWS Management Console now displays the account name directly in the navigation bar. I jump between so many AWS accounts every day that the reduction in stress from this seemingly simple update might actually add years to my life.

On the AI front, Amazon Bedrock continues to level up. It now supports 1-hour prompt caching, which should dramatically help reduce both latency and cost for longer-running or iterative workflows. Those of you building agents also get two meaningful upgrades: the AgentCore Browser now supports custom browser extensions (what could possibly go wrong?), and Bedrock now supports server-side custom tools via the Responses API. Being able to execute tools directly without orchestrating them through the client could be a huge unlock.

There were also some awesome serverless updates. AWS Lambda now supports .NET 10, continuing the fast pace of runtime updates. Cross-account access for DynamoDB Streams finally removes some long-standing friction for multi-account event pipelines. And payload sizes are expanding across the ecosystem, including Amazon EventBridge now supporting 1 MB event payloads.

Plus, Amazon Cognito introduced inbound federation Lambda triggers, allowing more customization during authentication flows. And Amazon ECR now supports cross-repository layer sharing which improves push performance and reduces duplicate storage. I've been waiting for this one to make managing Lambda base container images easier (and cheaper).

Unfortunately, there was some sad news this past week. Amazon laid off 16,000 workers, including some of my favorite people there. Cost shifting to chase AI may sound like a good short term strategy, but I fear that all these papercuts are going to eventually do some real damage.

In the Serverless Database Space Race, Databricks is turning some heads with its new Lakebase serverless database. Separating storage and compute isn't anything new, but using a Postgres interface to write directly to lakehouse storage in formats that Spark, Databricks SQL, and other analytics engines can immediately query without ETL is huge.

And finally, Astro is officially joining Cloudflare, tightening the loop between the modern frontend framework and the edge-first platform. In some ways I'm glad that these frameworks are getting the long term support they need. But on the other hand, the number of truly independent, open-source frameworks seems to be waning.

Tutorials

• CDK - Using Central Register Pattern for Resource Sharing by Tycko Franklin
• Migrating a Lambda to a Durable Lambda Function
• Building Event-Driven Observability on AWS Serverless (Part 1) by Sourabh Choubey
• Building Serverless Microservices on AWS with ECS Fargate, ECR, and Terraform by Kevin Kiruri
• Securing Serverless APIs with Amazon Cognito and API Gateway JWT Authorizers by Geovane Oliveira
• Mastering Serverless Data Pipelines: AWS Step Functions Best Practices for 2026 by Jubin Soni
• Building a Production-Ready Text-to-Text API with AWS Bedrock, Lambda & API Gateway by Oteng Isaac
• Using the VSCode Claude Code Extension with Bedrock and Claude Sonnet 4.5 by Matt Bacchi
• Strands TypeScript SDK - Building Production AI Agents by Martin Mueller
• Building an MCP Server on AWS Lambda: Complete Serverless Architecture Guide by Lucas Geovani Castro Brogni
• Amazon Bedrock Guardrails - Step-by-step implementation with Serverless by Walter Fernández
• Scaling Amazon Kinesis and AWS Lambda by Ankit malik
• Building an end-to-end insurance claims app with Amazon Bedrock by Arpad Toth

Reads

Do we need AWS Durable Functions when we have Step Functions?
Great post by John Nguyen comparing AWS Lambda Durable Functions with Step Functions. I'm low-key excited about the simpler code-based approach here (you know how I feel about bifurcation of business logic), and even though I wished they followed an existing syntax, it's still worth exploring.

I Migrated 40 Lambdas to Containers. AWS Bill Went Down 73%.
Another interesting post on the economics of serverless. Obviously AWS needs to mark up Lambda functions because you're paying for convenience and operational simplicity. That's often well worth it, but not always.

The Data Liberation: Amazon Athena and the Architecting of a Serverless Future
Athena's an absolute beast, and its API-driven approach fundamentally changes how we approach data analytics. The schema-on-read capability lets you analyze raw S3 data without ETL pipelines, and the zero-floor cost model really does give startups close to the same analytical power as enterprises without a costly Snowflake bill.

Weighing the benefits of AWS Lambda's durable functions
David Linthicum breaks down how durable functions work and what they mean for serverless adoption. He highlights the native state management capabilities and the workflow orchestration improvements, but also the complexity that comes with debugging these very AWS-specific abstractions.

AI-Driven SDLC: How to Build Secure, Governed, and Scalable Software with AI
Ran Isenberg explains how spec-driven development can tame the chaos of AI-assisted coding. I agree that writing structured specs before letting AI generate code prevents some of the slop we're seeing from yolo vibe coders, but I'm still not convinced that this becomes the standard approach.

When Serverless is MORE Expensive: 5 Architecture Patterns That Should Use ECS Instead
Dinesh Kumar Elumalai walks through five architecture patterns where ECS Fargate delivers better economics than Lambda. I've been known to make these same arguments. It's all about picking the right tool for your workload, and thanks to Dinesh, you now have a handy decision framework.

AWS re:Invent 2025 Deep Dive: The Truth About Lambda and S3's Facelift
A rather critical technical analysis of the AWS re:Invent 2025 announcements. It focuses on Lambda and S3 enhancements, why some are good, why some are bad. Lots of valid observations in here worth considering when choosing your next service.

Podcasts, Videos, and more

App Modernization with CDK Blueprints | Serverless Office Hours
Fabrianne Effendi and Jan Tan join Julian Wood to walk through Application Modernization CDK Blueprints. They cover composable patterns for intelligent document processing, agentic AI workflows, and full-stack applications. What a time to be alive. 😉

AWS Bites #151: EC2 ❤️ Lambda - Lambda Managed Instances
Eoin and Luciano explore Lambda Managed Instances with a practical video-processing demo using Step Functions. Scale up and scale down addresses real pain points, and I think there are definitely practical use cases for steady state serverless apps. It definitely blurs the serverless line, but that's a battle we lost long ago.

New: AWS Lambda tenant isolation | Serverless Office Hours
Really nice demo by Anton Aleksandrov and Ayush Kulkarni on Lambda's new tenant isolation mode. The automatic tenant identifier logging alone is worth checking out, especially if you've ever tried debugging which tenant triggered an issue in a multi-tenant setup.

What's New: AWS Lambda event source mappings | Serverless Office Hours
Rohan Mehta and Nihar Sheth join Julian Wood to show off the latest Lambda event source mapping enhancements. Lots of very cool stuff, including a 90% cost reduction for low-throughput Kafka workloads and advanced batch controls for error handling.

Serverless Craic Ep. 79: Reflecting on The Value Flywheel Effect (5 Years On)
Great discussion by the Serverless Craic crew on why serverless-first thinking becomes even more powerful when combined with AI. The idea that context is now a first-class architectural concern is something I've been screaming from the top of my lungs lately, and they nail why it matters so much in the GenAI era.

New from AWS

• Amazon Lex launches improved speech recognition models for English
• AWS IAM Identity Center enables account access and application use in multiple AWS Regions
• AWS STS now supports validation of select identity provider specific claims from Google, GitHub, CircleCI and OCI
• Amazon RDS now supports IPv6 for VPC endpoints of RDS Service APIs
• Amazon Keyspaces (for Apache Cassandra) introduces pre-warming with WarmThroughput for your tables
• AWS Lambda launches enhanced observability for Kafka event source mappings
• Amazon ECS now publishes container health status as a CloudWatch metric
• AWS announces Deployment Agent SOPs in AWS MCP Server (preview)
• Amazon Route 53 Domains adds support for .ai, and other top-level domains
• Change the server-side encryption type of Amazon S3 objects
• AWS introduces additional policy details to access denied error messages
• Amazon EMR Serverless now supports AWS KMS customer managed keys for encrypting local disks
• Amazon RDS Blue/Green Deployments reduces downtime to under five seconds
• Amazon DynamoDB global tables with multi-Region strong consistency now supports application resiliency testing with AWS Fault Injection Service
• AWS Databases are now available on v0 by Vercel
• Amazon Redshift Serverless adds queue-based query resource management
• AWS expands Resource Control Policies support for Cognito and CloudWatch Logs
• Enhanced Transactions view now available in AWS Billing Console
• Amazon EMR Serverless adds support for job run level cost allocation
• Amazon Bedrock Reserved Tier available now for Claude Opus 4.5 and Haiku 4.5
• Announcing Apache Airflow 2.11 Support in Amazon Managed Workflows for Apache Airflow

Developer Tools

Lambdaliths are not only for APIs. They can handle events too.
Elias Brange introduces lambdalith, an npm library that routes asynchronous events from SQS, SNS, DynamoDB, and EventBridge through a single Lambda function. The library provides a typed API similar to Hono/Express, supports batch processing, middleware, and error handling.

I built a Serverless OpenAI Gateway to cut costs by 30% and sanitize PII (Open Source)
Interesting take on LLM cost optimization from Guilherme Ferreira. Using Cloudflare Workers for caching and PII sanitization is clever, especially when you consider the sub-50ms cache response times and practically zero cold starts on the edge.

A Strands Agent Template (For the Impatient)
Danielle Heberling created a GitHub template for deploying Strands agents to AWS Bedrock AgentCore Runtime. The template includes production-ready CDK infrastructure with IAM, logging, and tracing, plus local development setup and CI/CD with GitHub Actions.

Final Thoughts 🤔

Hey all! It's good to be back. Yes, I took another month off from the newsletter. I wasn't planning to, but life got in the way again and I decided to take my own advice and step back for a few weeks. Nothing to worry about, I'm still in good health, just a really busy January. Things are kinda sorta calming down a bit, so I'm hoping to get back to my weekly cadence starting…now.

Hope you're all well,
Jeremy

I hope you enjoyed this newsletter. We're always looking for ideas and feedback to make it better and more inclusive, so please feel free to reach out to me via Bluesky, LinkedIn, X, or email.