Latest Insights

Latest things that happened

A Kubernetes setup a small team can actually run
Latest Insights
A Kubernetes setup a small team can actually run

Kubernetes has a reputation for needing a dedicated team. For a lot of what people run on it, that reputation is earned by choice, not by the tool. A three-person team can operate a cluster comfortably if they decide early what they are not going to do. Let someone else run the control plane ...

By BrainSoft  |  5 03 September 2026
Pagination that stays fast on a large table
Latest Insights
Pagination that stays fast on a large table

A table with a few million rows and an ORDER BY created_at DESC LIMIT 20 OFFSET 400000 query looks fine in development. On production data the page takes four seconds, the database burns CPU, and the API starts timing out under load. The query is not slow because the table is big. It is slow because the database has to walk and throw away every row before th ...

By BrainSoft  |  5 18 September 2026
Getting machine learning past the proof of concept
Latest Insights
Getting machine learning past the proof of concept

A model that scores well in a notebook is a good sign. It is not a product. The distance between "the metrics look right on my machine" and "this runs every day and someone trusts the output" is where most machine learning work quietly stops. The gap is rarely about the model. It is about everything around it: where the data comes from, wha ...

By BrainSoft  |  5 03 September 2026