Express Mode builds the ALB you skipped, and owns the knobs you used to write

Line up the two lists in the dev.to walkthrough and they match almost item for item. The author says the Terraform stack never got an Application Load Balancer, HTTPS, production-grade monitoring, alarms or scaling policies ^[7]. AWS says Express Mode creates a load balancer with listener rules, HTTPS configuration, auto-scaling policies, metrics, alarms and health checks on your behalf ^[3]. Each of the five gaps has a counterpart in what Express Mode provisions, with metrics standing in for monitoring ^[14]. That is a genuine result for a Flask portfolio site with three versions ^[11]. It is also why the comparison cannot answer the question a team with a running service is asking.

What is being traded away is not typing. AWS's own framing names load balancers, target groups and scaling policies as the manual configuration Express Mode removes ^[2]. Those are the objects whose settings decide how a deployment drains connections and when a service adds capacity. The walkthrough does not say whether the generated versions can be adjusted afterwards, or from where ^[13]. Anything the service stands up itself is, by construction, not a resource block an operator wrote and can read in a diff ^[12].

The shared balancer is the sharpest version of this. Up to 25 Express Mode services can sit behind one ALB, which AWS presents as a way to reduce the cost of running several services ^[5], and it is: the feature carries no charge of its own, so the balancer's hourly cost is the thing being spread ^[4]. It also means the listener rules your service depends on live on an object with as many as two dozen other tenants and no entry in your own change process ^[3][5].

The control point that remains is the input: a container image and an infrastructure role ^[2]. That role has to be permitted to create VPCs, subnets, security groups, balancers, alarms and scaling policies ^[3], which is not a permission set that scopes down neatly. The walkthrough runs on an IAM account with AdministratorAccess, and says outright that this is for the blog rather than production and that least privilege applies ^[8].

One edge is already visible. The author built for AMD64 and stayed there, noting that getting ARM64 working with Express Mode can be more complicated depending on the configuration ^[9]. On the traditional side, shipping a new version meant changing the Docker image in the task definition ^[10], and reaching the site at all meant running a script to find the task's public IP and appending port 5000 ^[6]. Neither stack here is one you would put in front of customers. The difference is that on the Express path, the parts you would need to fix are no longer yours to declare.