Good Morning!
Happy new year; we’re off to a slow start with some incremental product releases from AWS. Which frankly, I prefer over a bunch of stuff yeeted over the wall just as we’re getting back into the swing of January.
But first:
If you (well, not *you*, but probably the saddest looking person on your finance team) are tracking commitments in spreadsheets and hoping your discount strategy still makes sense, you’re not alone. Most teams are cobbling together strategies/tools that weren’t designed for the scale and complexity of modern cloud environments. That’s why we’re building Skyway over at [Duckbill](https://www.duckbillhq.com)—to take you away from all that. Now the exclusive sponsor of Last Week in AWS, and also the company I co-founded. Cloud contract issues? Get in touch.
Things I Found on the Internet
If your Tailscale network has grown beyond “just me and my Raspberry Pi” into actual infrastructure, you might like Tailsnitch. Automated security auditing for ACLs, node configurations, and all the stuff that’s easy to miss as your tailnet scales. Open source and actually useful.
AWS in 2026: The Year of Proving They Still Know How to Operate – I wrote this one, so I’m biased – but if you’re tired of breathless “is AWS dying?” takes, here’s my actual framework for 2026. The Azure numbers don’t mean what you think, Google Cloud is the real story, and re:Invent showed AWS finally accepting reality.
I wrote about this for The Register: AWS hiked GPU prices 15% on a Saturday, hoping nobody would notice. Spoiler alert: I noticed. The timing and method matter here – it’s the first real AWS price increase in years, and probably won’t be the last.
What AWS Has For Us This Time
AWS launches simplified import of CloudTrail Lake data in Amazon CloudWatch
“Simplified” here means “we finally let you move data between our own services without a Ph.D in data engineering.” It’s free to import, but CloudWatch’s custom logs pricing will make you wish it wasn’t. Check your wallet before consolidating anything.
Amazon ECS now supports tmpfs mounts on AWS Fargate and ECS Managed Instances
Hey look, tmpfs support finally made it to Fargate! Only took them… *checks calendar* …however many years since containers became a thing. What, half a century or so? Great for ephemeral data that doesn’t need to survive your next invoice surprise. At least they’re catching up to features Docker had in 2015.
EC2 Capacity Manager now includes Spot interruption metrics
Finally, metrics for something we’ve been flying blind on for years. Now you can quantify exactly how often AWS yanks the rug out from under your workloads. Free feature, sure, but mostly because charging for interruption visibility would be too on-brand even for them.
.NET 10 runtime now available in AWS Lambda
Microsoft ships .NET 10 three weeks ago, and AWS already has runtime support? Someone clearly learned from the .NET 8 debacle. Of course, it defaults to Native AOT, which means you’ll discover your dependencies aren’t compatible right around deployment time. Good luck explaining that delay to your PM.
Twenty percent cost savings sounds impressive until you realize it’s because AWS was charging you for overprovisioned local storage this whole time. Now they’ve graciously stopped billing for something that should’ve been serverless from day one. Progress, I guess?
Your AI Coding Assistants Will Overwhelm Your Delivery Pipeline: Here’s How to Prepare
So AWS is admitting their AI tools will drown your CI/CD pipeline in generated code you can’t possibly review or test fast enough. The solution? Buy more AWS services to automate around the problem their other services created. It’s the circle of (billing) life.
Build an AI-powered website assistant with Amazon Bedrock
They’ve reinvented the chatbot, but this time with enough AWS services to require a flowchart. Charging separately for the knowledge base, the LLM calls, the Lambda invocations, the ECS hosting, and the data transfer means that your support costs just got “scalable.”
… and that’s what happened Last Week in AWS.
