Good Morning!
And we’re met with the detritus of the stuff that didn’t quiiiiite make the re:Invent launch window.
From the Community
AWS gave a 300-level re:Invent talk about their outage in October, and I declare victory because internally DynamoDB apparently uses Route 53 as a database. I TOLD YOU!
HashiCorp’s fancy way of saying "nobody wanted to write TypeScript when HCL worked fine" just torpedoed thousands of hours of developer investment. But sure, "product-market fit" sounds better than "we made a thing nobody asked for." Except people are pissed that they EOL’d terraform-cdk with no notice.
My re:Invent talks have made it to YouTube. Check out What’s New with AWS Cost Management (COP203) – YouTube and Disagree in Commits:The Performance Improvements That Cut Costs by a Third-OPN309
Chris Farris seems bitter about missing re:Invent (I know I missed his presence), but at least Portugal has better port wine than Vegas has watered-down conference drinks for 10x the price. The piece I missed: AWS quietly making S3 Block Public Access enforceable at the org level while charging $110/month per VPC for encryption "compliance theater."
My review of the last ten minutes of Matt Garman’s keynote has been published in The Register.
Podcasts
Last Week In AWS: Corey Quinn Crashes Out
Screaming in the Cloud: The AI Productivity Gap with Keith Townsend
Choice Cuts
Exploring the new AWS European Sovereign Cloud: Sovereign Reference Framework – AWS builds a separate European cloud because GDPR scared American companies, then charges premium prices for what amounts to "we promise your data stays in Europe and Europeans run it." What does this solve for? Nothing other than "regulatory checkboxes."
Now generally available: Amazon EC2 C8gb instances – Graviton4 chips deliver 30% better performance, which is impressive until you remember the last three generations also promised 30% improvements. At some point the math stops mathing, or AWS is admitting their older chips were embarrassingly slow.
Amazon CloudWatch SDK supports optimized JSON, CBOR protocols – AWS finally upgraded CloudWatch’s protocol from something that predates the iPhone. The performance improvements are real, but calling this a feature announcement is like bragging about switching from dial-up to broadband in 2025. At least it’s free—for now.
Building national foundation models – Nations rushing to build sovereign AI while AWS quietly becomes the world’s landlord. Turns out "digital independence" means renting GPUs from Seattle, but with extra steps and a patriotic press release. At least the invoice will be in your local currency.
New report: Cloud “fundamental” for European national security and defense – AWS funded a think tank to write a report concluding that governments desperately need to buy more AWS. They’re probably right about legacy systems being garbage, but framing vendor lock-in as "strategic readiness" takes serious chutzpah.
AI Increased Productivity? Consider Hiring More Developers! – And with the same energy, AWS just published a blog arguing companies should hire more developers because AI makes them productive, which is like a car dealership saying you need two cars since they’re so fuel-efficient now. The logic isn’t wrong, but watching AWS tell CFOs to increase headcount while selling productivity tools is peak irony.
IAM Policy Autopilot: An open-source tool that brings IAM policy expertise to builders and AI coding assistants – AWS just open-sourced a tool that writes IAM policies so developers can stop pretending they understand them. It generates overly permissive baseline policies that "prioritize functionality over minimal permissions"—which is exactly how most production policies end up anyway, except now you can blame the robot instead of your hasty deadline. IAM is going to be the death of this platform.
AWS and Google Cloud collaborate to simplify multicloud networking – Again, this is either wonderful or terrible depending entirely upon how much it’s gonna cost, but they aren’t saying.
Exploring Optimize CPU feature on Amazon RDS for SQL Server – SQL Server licensing costs so much that AWS built an entire feature to help you use fewer cores. The real trick here is helping you avoid explaining to your CFO why the database costs more than the entire engineering team’s salaries combined.
Prometheus MCP Server: AI-Driven Monitoring Intelligence for AWS Users – They’ve solved the "PromQL is hard" problem by adding an LLM intermediary that costs money to run and occasionally hallucinates metrics. "Hey ChatGPT, what’s on fire?" "You’re absolutely right!"
Tools
Friends don’t let friends use Control Tower, which wouldn’t help anyway for multi-organization management. I’m keeping an eye on the just-launched Quiverstone for that. It’s early, but promising.
… and that’s what happened Last Week in AWS.
