🦸 Community Superhero
Our community superhero this week is Prashanth HN, co-founder and CTO at AntStack and AWS serverless hero. Prashanth is an brilliant technical mind and is wonderful at explaining complex ideas in easy-to-understand ways. You can find him all over the world talking about serverless, AI, and modern cloud development. Thank you for everything you do, Prashanth!
💯 Spotlight
By now many of us have tried our hand at AI agent building. Just like any production code, there’s a “fun” way to do it and a “right” way. But let me throw a spin on it that you might not have thought about. What about voice agents? The robots you talk to on the phone when you call a support line or on an Alexa-like device fall under much tighter constraints and harder requirements than you’d expect, especially compared to behind-the-scenes autonomous agents. The team at Dograh wrote an eye-opening post about all the production-ready constraints this modality falls under. It has some vendor pitching in there, but the big takeaway is the practical considerations to building quick-response agents. Very interesting read!
🔥 My Favorite Content
I clearly missed the Ralph meme in the AI world. Tbh I’m not entirely sure what it means, but Malte Ubl hooked me with it in a blog post last week stating that Vercel “Ralph Wiggumed WebStreams to make them 10x faster.” As funny as the title is, the content is a performance engineer’s dream. It goes through the steps they took to make webstreams orders of magnitude faster and how they patched it so you have to change as little code as possible. One of my biggest takeaways though is that they rewrote this primarily with AI and used the massive test suite to verify feature parity. Wonderful article with great lessons and technical depth.
That said, not everything needs AI. As easy as it is to toss an LLM in the middle of a workflow to do something, we need to start pausing and asking ourselves if it could be handled in plain “old-school code.” We got a solid reminder of this with a blog post from Sandeep Mishra, Ali Syed, and Santosh Jade last week where they walked us through a simple-yet-elegant build that automatically decompresses files in S3 with Step Functions. This serverless workflow checks the size of an incoming file to S3 and either unzips it in Lambda or a spun up/spun down EC2 instance. This is a highly-repetitive task that would be easy to give to an LLM, but really isn’t necessary and would be a waste of inference. Check yourselves, people!
For the app builders, Darryl Ruggles gave us a great reference architecture and explanation of Lambda Durable Functions. In his post, he builds a loan approval workflow that pulls out all the stops - waiting for callbacks, checkpoints, idempotency checks, and quite a bit more. While the code samples he shares are useful, I found the most value in his insights, like when to use Durable Functions over Step Functions and things to watch out for.
Staying on the Durable Function train, Daniele Frasca wrote a comparison blog where he built the same workflow in both Durable Functions and Step Functions. This post is much less about gotchas and more comparing the DX and cost between the two. This is a great post on its own, but really shines when read along with the one from Darryl. I love that both of them approached this from different angles, it builds a great picture for app builders.
💡 Tip of the Week
There’s no doubt coding agents like Claude Code are reshaping how we do development work. Lenny Rachitsky gave us a breakdown of what he took away from Boris Cherny with respect to coding. These points feel like they’re a bit ahead even of the early adopters, and I also feel like they need to be taken with a grain of salt. Great food for thought nonetheless.
🐣 New Releases
Reminder, all releases from AWS can be found on AWS News by Luc van Donkersgoed. Below are my favorite from last week.
Anthropic released Sonnet 4.6 on Tuesday and it’s already made its way into Amazon Bedrock and Kiro. My experience with it so far has been wonderful - particularly in Claude Desktop. Give it a try if you haven’t already.
Aurora DSQL got a big boost with integration into Kiro powers and AI agent skills. They can now help you build apps backed by DSQL via schema design, performance optimizations, and supported database operations. Nice!
DSQL also launched Connectors for Go, Python, and Node that simplify IAM auth if you’re using standard Postgres drivers to connect. They handle all the IAM auth for you in the libraries you already know and love. Big win!
Last Words
I don’t know about where you live, but here in Texas the sun is out and the grass is starting to turn green! Spring energy is in the air and I’m excited to see what comes in the next few months as we shake the cold and get remotivated.
That’s my take on the week, but what’s yours?
What did I miss? What made you nod along (or 🙄)? Hit reply if you’re reading the email. Prefer socials? Ping me on Twitter, LinkedIn, or email.
Happy coding!
Allen