Writing for engineers
who care about the
inside of the machine.
Long-form essays, postmortems, and field notes on systems, performance, AI, and the discipline of shipping.
The Living Dead: Why Java Feels So Exhausting Now
After another night battling a production incident fueled by Java's 'enterprise-grade' complexity, it's time to admit: Java isn't dead, it's the undead zombie in our server rooms, a costly and exhausting relic.
RabbitMQ vs. Kafka: When the Diagrams Lie and Prod Explodes
Another 3 AM call, another 'why is this broken?' moment. Let's talk about RabbitMQ and Kafka not from some clean architecture diagram, but from the trenches where message queues and event streams either save your ass or become the reason you're looking for a new job.
The Notification Grind: Go, Node, and RabbitMQ in Multi-Tenant Hell
Remember that Tuesday at 2 AM? The one where a minor tenant's custom webhook brought down notification delivery for everyone? Yeah. We ended up deep in the trenches, comparing Go and Node.js for a critical RabbitMQ-backed service.
Cutting Through the Noise: A Late-Night Rant on Directness in Systems
Another 3 AM production incident survived. Time to talk about why we make our systems so damn complicated, and why sometimes, the most elegant solution is the one that just gets straight to the point.
AI Coding Assistants Fail at Production Debugging
AI coding assistants generate correct-looking code but often fail in production debugging. Learn why runtime profiling, system constraints, and execution paths matter more than generated solutions.
From 500ms to 900ms: How AI-Assisted “Optimizations” Turned a Fast Query into a Slow One — and What Brought It Back to 43ms
An API endpoint went from 500ms to 900ms after AI-suggested “optimizations,” until removing ORM abstraction and switching to raw SQL reduced it to 43ms, revealing how performance depends more on system understanding than generated fixes.
It worked on localhost. Obviously.
Localhost is a comfortable, consequence-free fantasy. Production is where software goes to find out what it actually is. A 3am field report from someone who has been profiling things nobody asked them to profile.