Domain knowledge + System level thinking = Enterprise level software.

One amazing thing about building AI is that you spend relatively much less time thinking about implementation and more time thinking about selling and growing the product.

You’re not thinking about how to get this feature done, about how you lack language or stack proficiency

So many yaml files in this project I don’t know which is doing what and have to refresh my knowledge each time 😆

I use Jetbrains Junie for tasks, but a frustration I have with it given that I’m building with a stack I’m not familiar with is that its explanations are poor and misses important details sometimes

This is where I switch to Copilot (Sonnet 4) in Agent mode (no code changes), it’s a great teacher

Today I implemented mTLS encryption and JWT for secure service to service communication and authorization.

I made use of OPA as the policy agent and authorization decision point, then used Envoy Proxy as the sidecar for policy enforcement in my Kubernetes cluster

Past few weeks has been about setting up and configurations

My project is a Golang, React and KMP monorepo basically

Backend services would be Go, Web Frontend would be React, Mobile apps KMP. Infrastructure would be Docker compose + Kind for local dev, Terraform + Helm charts, AWS EKS for prod

Building software alone is hard, especially when it’s all new and it’s just you and an LLM dev tool.

So I’ll just make this my standup space and share as I build. Hopefully this keeps me going and I don’t end up abandoning this project like every other one I started