For those of you looking to lose some weight in 2026:
- Join a fitness class
- Walk as much as possible (invest in a walking pad)
- Track your calories (honestly)
Give it some time. You’ll get there.
Kubernetes is a solution, not a default.
But somewhere along the way, it became the answer before anyone asked the question.
Teams pick Kubernetes for three bad reasons.
- Everyone else uses it.
- It looks good on resumes.
- "We'll need it when we scale."
The last one sounds the most reasonable and does the most damage. It front-loads years of operational cost against growth that may never require it.
Here's what Kubernetes actually solves. It schedules many workloads across a shared pool of machines. It gives dozens of teams one deployment model, one API, one way to handle config, secrets, scaling, and service discovery.
Notice what those have in common. They're problems of scale and heterogeneity. Many services, many teams, many workload types.
If you don't have those problems, Kubernetes doesn't solve anything for you. It just adds a layer.
And the layer is expensive.
The EKS control plane costs $73 a month. That's the cheapest part of the bill.
- A minimal production setup needs a load balancer controller, external DNS, secrets management, metrics server, node autoscaler, monitoring stack, and ArgoCD. Seven components before your application exists.
- Each one has its own release cycle, CRDs, breaking changes, and IAM role.
And you're not upgrading one thing. You're upgrading a stack. Every four months.
My rule of thumb: running production Kubernetes well takes at least half of one experienced engineer's time, permanently.
For a ten-person engineering org, that's 5% of your total capacity spent on the platform rather than on the product.
For most teams, ECS on Fargate is the answer. Containers without nodes, upgrades, or add-ons. Secrets, logging, autoscaling, and deployment rollbacks built in. AWS patches everything underneath.
Kubernetes starts making sense when you have 20+ services owned by several teams, complex scheduling needs, or workloads that ECS simply can't handle.
What makes adopting Kubernetes easy later isn't adopting it now. It's containerized services, stateless processes, and infrastructure as code.
A team on ECS with clean containers can migrate to EKS within a few weeks as needed.
Choosing the simpler option isn't the cautious move.
It's the one that keeps your options open.
and please don’t give me any props lmao. you all will notice i tend to do this once a year at this point.
but once i hit month 4 and month 9, i will need yall to celebrate with me. especially after the 1 year mark.
black people really grind my f***ing gears sometimes lmao. this is why i say no to gigs i don’t wanna do and why i have a 9-5. im not playin wit mfers no mo
cause i don’t blame people for not wanting to learn Git, Docker, Terraform, Ansible, Jenkins, Linux, and Kubernetes. that’s a lot.
leave it to me tho. I’ll do it