I've Been Using AWS Wrong for YEARS...

For years, my approach to AWS felt like a battle.
As a DevOps engineer and later and architect, building infra always involved a tedious process of carefully building templates and structure, reviewing, deploying, testing and iterating over and over.

I’d either spend hours clicking through the console or writing endless infrastructure code, always feeling like I was one misconfiguration away from a headache.
It turns out, I was making it much harder than it needed to be by not using the right kind of tools for the job.

The biggest problem we face as developers isn't writing application code, it's the gap between a working project on our local env and a secure, scalable, and cost-efficient setup in the cloud.

DevOps is complex, and while general-purpose AI chatbots are great for generating a function or explaining a concept, they often fall apart when you ask them to build real production infrastructure. They lack the deep, specialized context and can hallucinate configurations (or, misconfigurations rather) that look plausible but are practically broken.

This leads to a frustrating loop of generating code, trying to deploy it, watching it fail, and then manually debugging what went wrong.

I recently decided to built my own Home Barista project so that guests could order drinks using an app (yea, I know, save it 🤣).
I wanted to put this project in production, and while I have an endless amount of options, I'm also the proud owner of $600 credits on AWS, valid till next year...

What I learned from building that project is that the solution isn't just any AI, but a specialized one that can speak the language of the tools we use.
The game changer for me was the Q Developer CLI.
Yes, it's AWS's, but also, completely open source and free to use (you are after all building infra on their cloud...)
It’s so deeply trained on AWS that it can offer surprisingly sharp insights.
For fun, I asked it to roast the AWS dashboard, known for it's "annoying" UI, to put it mildly, and its response was spot on:

The real magic, however, comes from their MCPs.
Now, I know, I know you've heard the buzzword so many times you're almost clicking away, but stay with me.
​AWS came up with their own long list of MCP servers free and available to use, that I've plugged into Q to get help with both terraform code (trained on AWS official Terraform provider) + their cost analysis MCP helping you drill down in cost allocation, reduction and other wasted infra on AWS.

My workflow looked completely different.
I started with my application code and a simple architecture diagram.

Then, I fed my architecture to the Q CLI.

I asked it to build the entire production infrastructure using its Terraform MCP.
It generated all the necessary files, creating a robust setup with networking, container services (ECS), and databases.
​

Here’s the incredible part

I then asked it to use its Cost Analysis MCP to tell me how much the infrastructure it just designed would cost per month. It came back with an estimate of around $100 ("way too bloated if you ask me", I thought).
So, I simply told it: "That's too expensive. We don't need two availability zones or a load balancer."
Q understood, updated all the Terraform files, and provided a new cost breakdown $45/month (still not the $5 you can get elsewhere, but imagine a full blown networking setup, load balancers, monitoring and everything around it, ready for production grade scale).
Oh yeah, and I have these credits waiting 😉

The best part? The generated Terraform code worked on the first try.
terraform plan and terraform apply ran smoothly because the AI wasn't just guessing; it was using the official Terraform provider for AWS, ensuring all the resources were configured and connected correctly.

I don't know if you've ever built something slightly complex on Terraform, but there's one rule, it never, ever, runs on first try. Mine did!

This approach flips the script.
Instead of using AI to generate disconnected snippets that I have to stitch together, I'm using a specialized partner to architect, build, and cost-optimize an entire system, all from my command line. It's a faster, smarter, and frankly, more enjoyable way to work with the cloud.

And while the above is just an example, imagine taking this process to your work environment, cost reduction projects and enterprise-grade systems. It's a game changer!

​
Thank you for reading.
Feel free to reply directly with any question or feedback.
Have a great weekend!