All you need is PostgreSQL

There's a quiet rebellion brewing in software engineering, and it starts with a simple provocation: stop adding tools you don't need.

A new technical deep-dive argues that most modern engineering teams reflexively pile infrastructure on top of infrastructure — caching layers, event-sourcing databases, specialized analytics engines — when PostgreSQL alone could handle the job.

Using a real-world financial transaction system as a test case, the piece walks through how vanilla PostgreSQL 18 can enforce complex business rules, maintain full audit histories, handle both transactional and analytical workloads, and scale comfortably to mid-sized bank volumes, all without distributed systems acrobatics.

The core insight is pointed: complexity grows not because problems demand it, but because someone saw a tool at a conference talk.

For startups processing around 10,000 transfers daily, the estimated working memory footprint is a modest 22 megabytes. Even at the scale of a major bank like Nubank, PostgreSQL running on current AWS infrastructure could theoretically hold the entire working dataset in memory.

Sometimes, the boring choice is the brilliant one.