Our serverless superhero this week is James Ward, principal developer advocate at AWS, book author, and podcast host. James is a fantastic educator on all things serverless and AI. The way he talks about GenAI and serverless is so relatable and fun, it’s always a joy to learn from him. If you ever get a chance to listen to him talk at a conference or hop onto his podcast, do it! You’ll learn lots and have a great time doing it. Thanks for everything, James!
I read an article from Marc Bowes last week that was immaculately written. The post, DSQL sucks at TPC-B, is in response to an article comparing DSQL to Aurora Serverless v2. I went into it thinking it was going to be a clapback, but Marc remained composed the entire time, offering reasonable explanations for what the author saw and did his best to offer alternatives. This is an extremely technical write-up (and honestly a little bit over my head) and I love it. DSQL is a modern feat of engineering and it’s really cool to see when the people who built it step up and show you how it’s meant to be used and tuned. Amazing.
In another high-energy-and-extremely-information-dense video, James Eastham shared with us last week a question and answer on whether you should emulate the cloud locally on your machine. Developers just getting into the cloud after years of on-prem experience tend to have different answers on this than developers who were born into the cloud. James gives an answer I didn’t quite expect, but backs it with pretty fair reasoning and proof. Thanks for making us think outside the box, James!
Seth Orell wrapped up his 5-part series on How to Crater Your Database last week with a summary of all the conclusions he’s made over the past several weeks. I really liked this series - it gets pretty technical, but explains concepts in a relatable way that make them easy to understand. Given the outcomes of the other articles in his series, his summary was not a surprise but a good reminder that performance in any database at scale is your responsibility. You can’t just throw bad design at a wall and hope it sticks.
We’re currently in the wild west of AI development. I say that generically across all meanings of that phrase. We’re exploring. Tech is changing. People are just now starting to understand use cases that AI can actually solve. I read a sobering article on the percentage of tasks AI agents are currently failing last week that predicted doom and gloom. But to me, it sounds like early adoption. Early adoption with the biggest spotlight we’ve ever seen. I still remain optimistic on the practicality of AI agents, but I feel like most people aren’t quite there when we’re talking about how to build with them meaningfully and securely.
I’ve been working really hard lately to make the A2A protocol easier to use and serverless compatible. My first iteration got it working in Cloudflare Workers. Last week I released my second iteration, which makes it fully compatible in Lambda with Amazon Bedrock. It greatly simplifies the process for building agents that can communicate with each other 🔥
Reminder - AWS News is the best source for AWS-related service announcements. For all releases and summaries of what happened, head over there!
DynamoDB global tables with strong consistency when generally available last week. Perfect for global applications with strong consistency requirements.
Amazon Nova Canvas got a cool update last week. It now lets you do a virtual try-on and adapt styles of images. So you can ask it to see what you’d look like wearing certain clothes! The blog post I linked shows how it works, it’s pretty neat.
Amazon S3 Express One Zone now supports tagging for both cost allocation and for ABAC. These are huge features for production-ready systems.
For the Americans - happy belated Independence Day 🎆! I hope y’all had a good long weekend and are feeling refreshed and ready to build something new. If you want to collaborate on something, I’d love to hear about it!
If you’d like to make a recommendation for the serverless superhero or for an article you found especially useful, send me a message on Twitter, LinkedIn, or email.
Happy coding!
Allen
Thank you for subscribing!
View past issues.