🦸 Serverless Superhero

Our serverless superhero this week is Kyler Middleton, Microsoft MVP, AWS Community Builder, podcast host, and principal developer at Veradigm. Kyler does fantastic work in the world of AI and serverless, including writing a fully comprehensive guide on building a Slack bot with Lambda and Bedrock. She is a wonderful, friendly teacher with a firm grasp on the concepts she’s teaching. Thank you for everything you do, Kyler 💙

💯 Spotlight

There’s never been a bigger push on workflow orchestration than right now. With AI agents on the rise, people are pushing the boundaries of what we can “just do.” But we can’t forget the basics—and that’s exactly what Matt Morgan reminded us of last week with his AWS Step Functions and JSONata blog post. Somebody has to build the workflows! So why not make it markedly easier on yourself and try out the massive simplification that came with JSONata and variables.

🔥 My Favorite Content

There’s so mch power behind cautionary tales. When you read them, you can tell there’s a certain angst present that you don’t get with theoretical posts. Matt Gillard wrote an article last week with lots of heart, warning us about Amazon Data Firehose to S3 Tables. I doubt what he’s telling us about in that article is documented, so I’d be sure to read what he has to say before starting down that path.

Speaking of cautionary tales, I had one last week, albeit much more on theory than practice compared to Matt. As I’m exploring the model context protocol (MCP), I’m watching patterns emerge and am fearful of what’s happening. We’re moving so fast that some things appear to be completely missed or ignored, like stealing someone else’s API to use in your agent tooling. I proposed something similar to CORS for AI agents to prevent unauthorized use of APIs. I’m curious what y’all think.

Despite the misuse of APIs enabled by MCP, everyone is going all-in right now. We received a handful of MCP servers from AWS last week, signalling an even broader adoption of the protocol. We also are seeing cool “unofficial” MCP servers for AWS, like the AWS Service Reference Information MCP server from Paul Santus.

I love seeing solo projects from builders in the community. Pubudu Jayawardana shared a project he made for a hackathon back in December where he used Amazon Bedrock and Step Functions to build a simple spelling game. It’s a good reminder that great software doesn’t need 1000 moving parts. There’s elegance in simplicity, and that’s exactly what we’re seeing from Pubudu’s project. He shares his lessons learned, the back and forth he had with Amazon Q during the build, and of course, a link to play the game yourself.

Yan Cui brought back a great discussion on the pros and cons of the lambdalith. Lambdalith refers to using a single Lambda function as a monolith—like running all your API endpoints in a single function. He gives the pros and cons of building this way in a well-thought-out discussion. I like his thought process and completeness on the subject. But with everything in computer science, the right answer on whether it is the best decision for you is “it depends.”

💡 Tip of the Week

Despite all the genuinely transformative features we’ve been seeing with AI, it’s still easy to get lost and mess stuff up - especially now that computer access is a thing. This post from Danny Thompson is a great reminder that with great power comes great responsibility.

🐣 New Releases

Amazon CloudWatch logs increased the max log event size to 1MB, up from 256KB. I can see a few use cases for this, but be wary of how much this will affect your monthly bill!

AWS App Studio now offers a prebuilt solutions catalog, which sounds to me like blueprints of apps with best practices built in. Nice!

Amazon API Gateway now supports IPV6 for all endpoint types. That’s a big one!

Last Words

Happy birthday to the newsletter! This issue marks 3 years of the newsletter without missing a single issue. My sincere thank you to all of you for subscribing and for the community members who helped keep the streak alive during my family emergencies last year.

It’s getting increasingly more difficult to not talk about AI. It’s got a solid foothold in software development both in how we build and what we build. What are your thoughts? Is it too much? Not enough? What are you most excited about?

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