Writing

Notes on Node, Kafka, Postgres, and the parts that turned out harder than they looked.

2026.09.02Why your Kafka consumer lagsConsumer lag is usually not a throughput problem. How to read the number, the five causes that actually produce it, and how to tell healthy lag from a consumer that is falling behind.kafkanodejsperformance7 min read

2026.08.14The Postgres indexes you're not usingPartial, covering, expression, and BRIN indexes solve problems a plain B-tree can't — and pg_stat_user_indexes tells you which of your existing indexes are dead weight.postgresperformancesql6 min read

2026.07.30GraphQL N+1, actually solvedDataLoader is the answer to GraphQL's N+1 problem — but only if you scope it per request, keep the batch contract, and know the one case it can't handle.graphqlnodejsperformance5 min read

2026.07.11Strapi in production: what the docs don't tell youSchema as code, media that survives a redeploy, queries that don't populate the world, and keeping slow work out of lifecycle hooks — running Strapi for real products.strapinodejsdevops5 min read