Coding agents pile up history: every command, every file read, every test log. A recent paper, You don’t need long context, shows that frontier models lose almost nothing when they compact that his...
The previous two posts in this series covered token bucket basics and scaling to 1.7M ops/sec. Both stayed inside one algorithm family. This post goes wider: GCRA and Sliding Window Counter — two i...
Last week I benchmarked three token-bucket implementations and crowned Lua+EvalSha the winner at ~66K ops/sec on a single Redis node. That was the floor. This post walks from there to 1.7M ops/sec ...
Implementing a distributed rate limiter with Redis? Three approaches: Lua scripts, multi-command client-side calculation, and WATCH optimistic locking. I benchmarked all three. Here’s what I found....
The hardest problem in device communication over MQTT is not delivery. It is semantic consistency under concurrency. Once you scale beyond a single service instance and a single stable device conne...
Every time someone asks why closing a channel twice panics, the real question is usually about ownership. A channel has exactly two states: open and closed. close(ch) flips a bit in the runtime ...
I recently spent some focused time reading PostgreSQL’s concurrency and locking documentation, and then mapping that understanding to how locks actually work in Spring JPA / Hibernate. This is not...
Needed to share my local dev server externally. LocalTunnel made it instant. What It Does LocalTunnel exposes your localhost to the internet with a public HTTPS URL. No config, no account need...
Had Bluetooth audio dropouts with my Sony speaker. Switching from AAC to SBC fixed it. The Problem Sony Bravia Theater U speaker kept stuttering and dropping connection on macOS. Audio would cu...
Switched Emacs Org mode LaTeX previews from ImageMagick to dvisvgm. Much sharper results. The Problem Default LaTeX preview in Org mode uses ImageMagick to convert equations to PNG. Results are...