Deployments, Gen-AI, and How They Shape My Development Process

Hey, my name is Quentin Mayo, and in today’s blog I want to talk about deployments, Gen-AI, and how both shape the way I think about development. I’ll start with a story—because that’s how my brain usually connects everything.

Why I “Never Look Back” in Development

There’s a moment in The Incredibles where Edna Mode says she never looks back. That mindset stuck with me. I treat my devices and development setups the same way. Every so often, I’ll do a full hard reset on my phone or laptop. I back up what I know I’ll need… and if I didn’t grab it, it’s gone forever. No regrets.

I apply that same philosophy to my development environments. Throughout my home lab and even in professional work, there have been countless times where I built something—a pipeline, an application, or a deployment setup—then realized I could build it better with the constraints I have today. Reset. Rebuild. Improve.

That’s really what this vlog is about.

My Deployment Evolution

When I first started building websites, apps, and just tinkering with entrepreneurial ideas, I followed the same path a lot of organizations take:

EC2 for basic servers
Then Dockerized containers
Then ECS as services needed to talk to one another
And exploring equivalents in other clouds

ECS was… complicated. Still fun, but definitely not simple.

Recently, I discovered tools like Coolify and Dockify—platforms that let you deploy multiple services on a single instance, with a clean UI. This changed everything.

Migrating WordPress

I moved all my WordPress sites:

From EC2
To Lightsail
And finally to Coolify

Lightsail was convenient… until it wasn’t. It wasn’t as cheap as advertised, had limitations, and I hit some scaling pain.

So I figured: why not run Coolify on-prem? I even got AT&T to give me a static IP (which they were not excited about), and it worked… until it didn’t.

Because I have a baby.
And that baby loves unplugging my servers.

One unplug → and boom → all my sites go down.

So now I’m moving toward a hybrid approach.

Designing a Hybrid Deployment Architecture

Here’s the concept:

Route 53 manages my DNS.
The primary A record points to my home network running Coolify.
If the home IP becomes unreachable, Route 53 fails over to:
→ an EC2 instance also running Coolify in the cloud.

This setup gives:

the low cost and flexibility of on-prem
the reliability of AWS
automatic DNS-level failover

Database Strategy

If I’m splitting workloads between home and cloud:

For small projects that don’t really matter → I’ll keep them on-prem.
For anything that needs reliability → the database cannot be on-prem.

That leaves two good options:

A database running on an EC2 instance (cheap, small projects)
RDS for anything larger, or anything requiring scaling or stability

A Common Mistake (AI Included)

One thing I’ve noticed—and AI models repeat this mistake constantly—is the advice to always put your database in a private subnet. Sounds good… until you realize your application might need direct access from outside, depending on your architecture.

Databases in a private subnet with no plan for safe exposure leads to:

misconfigurations
exposed IPs
unnecessary risks

You need a solid access pattern—not just “put it in private and pray.”

Gen-AI and Development Speed

Now for the Gen-AI side.

There are things I’m doing right now that would take some organizations weeks or months, maybe even a year or more. I built them in a weekend with Gen-AI assisting me:

Examples:

Setting up GitHub Actions with PATs and a GitHub App
Creating an autoscaling group of self-hosted GitHub runners
Reducing GitHub Actions billing by using my own metal before cloud minutes

It’s wild how much faster development is when you combine:

solid engineering fundamentals
strong problem-solving skills
and modern AI tooling

Gen-AI doesn’t replace thinking—it accelerates people who already know how to think.

Those who rely on AI without understanding fundamentals will struggle. But those who use AI to extend their reach will outperform by a mile.

Wrapping Up

I’m almost at my destination, so I’ll close here. Today’s thoughts were really about:

Why I rebuild systems instead of clinging to them
Why hybrid deployments make sense for my home lab
The realities of cloud cost vs on-prem cost
How Gen-AI lets engineers ship insanely fast—if they know what they’re doing

More to come soon.