Aider is my favorite AI coding assistant. It lives in the terminal, commits directly to Git, and actually understands my codebase. But it was missing one thing: the ability to save and restore chat sessions. So I built it. The PR never got merged, but I've been running my fork for months now. Here's the story.

Vector databases are everywhere now. Pinecone, Weaviate, Milvus - the list goes on. But what if you're already running ClickHouse and don't want yet another database to manage? Turns out, ClickHouse can do vectors too. This post explores what vector databases actually are, why ClickHouse might be a surprisingly good choice, and how to get semantic search running with plain SQL.

When thinking about performance, it's easy to focus on Big O notation. But in Go, the difference between the Stack and the Heap is often the difference between a service that scales and one that chokes on GC pauses. This post explores escape analysis, the "Pointer Myth", and why passing by value is often 40x faster than passing by pointer.

"The database that runs your app cannot be the database that analyzes your app". It's a hard lesson learned at scale. Early on, Postgres does it all. But as you hit massive scale, your analytics queries start killing your login APIs. This post breaks down the physics of Row-oriented (Couchbase) vs Column-oriented (ClickHouse) databases, and how to bridge them using Change Data Capture (CDC) for a robust, lag-free architecture.

"Why is our service slow?" "I don't know, the heap is only 200MB". "But we're allocating... wait, how much?" "12 terabytes". "...in how long?" "30 seconds profile". That's when we realized: we weren't running a service. We were running a garbage factory that occasionally served API requests. The Go garbage collector was heroically trying to clean up our mess, and we were blaming it for not cleaning fast enough. This deep dive into GC internals, profiling tools, and production war stories will teach you how to stop fighting the garbage collector and start working with it.

CouchLens is a browser-based tool for analyzing Couchbase N1QL query performance. It parses system tables, extracts execution plans, and generates insights to help database administrators find performance bottlenecks without sending data to external servers. This post explains what it does, how to use it, and what features are coming.

Working with Couchbase for the past year and a half has been more than just learning a database. It has been an exercise in system thinking: seeing how indexes, queries, consistency, and durability interact, and how small design choices ripple through performance and reliability. This essay is my personal notebook from that journey. It is technically dense because Couchbase demands precision, but it is also reflective because the lessons extend beyond one database. Understanding how array indexes multiply entries, why compound index order matters, or how query consistency flags change latency is not just about Couchbase, it is about learning to think in terms of systems, trade‑offs, and consequences.

Understanding Private-Public Key Encryption - The Backbone of Modern Security.

Learn how to refactor your code in 5 minutes.

My journey to build my own cloud storage from scratch with nextcloudflare and the lessons I learned along the way