Postgres Connection Pooling Under Sustained Load
At 02:40 the nightly reconciliation job stops running. Not crashing — stopping. The scheduler says it dispatched the run, the worker log says it acquired the...
Recent posts
Connecting Codex to Google Workspace via MCP
When you work with an AI coding agent like Codex, the real productivity gains come from letting it interact with the systems around your code — not just the ...
How a Language Model Learned to Speak Vietnamese
You’ve used Google Translate’s voice feature. You know the one — you type something, hit the speaker icon, and hear a voice that is technically correct but s...
Zero-Config Backends: The Architectural Future of TypeScript Frameworks
There’s a particular kind of frustration that only backend engineers know: you wrote the controller, you wrote the service, the types look great, the tests c...
Retries Without Thundering Herds
Retry budgets, jitter strategies, and queue backpressure for control‑plane operations
Predictive Scaling 2.0: Using Reinforcement Learning to Optimize Kubernetes Autoscaling Policies
It’s 02:37. Your phone vibrates like it’s trying to tunnel through the nightstand.
Hardening Server-Side React: Best Practices After React2Shell
React2Shell is the moment a lot of frontend teams discovered they actually run a server-side framework now.
From Kafka to NATS: When Less Is More in Distributed Messaging
You’ve got a Kafka cluster humming in production. There are topics, partitions, consumer groups, dashboards, and an ever-growing list of “we’ll tune that lat...
Bridging the Edge: Using NATS Leaf Nodes to Build Hybrid and Multi-Cloud Systems
If you’re running anything remotely serious today, your architecture probably isn’t “just” in one place.
Postgres as a Message Bus: Implementing Durable Event Queues with LISTEN/NOTIFY and Logical Decoding
You’ve got a small-ish system: a monolith plus a couple of background workers, maybe one or two sidecar services.
Beyond Postgres and DuckDB: The Rise of Composable Query Engines with Polars and DataFusion
If you’ve been doing data-heavy backend work over the last decade, your mental model probably looks something like this: