🫷Your Claude Code sessions get worse over time.
Not because the model is getting dumber. Because as the context window fills up, quality drops. Every file read. Every message sent. Every decision made. It all goes into one bucket. And when that bucket gets full, you start getting generic answers, forgotten instructions, and output that feels like it was written on autopilot. There's a name for it: context rot. Yesterday I was on an LGJ team call and heard about two tools built specifically to fix it. Didn't just take their word for it. Went straight to GitHub, checked every fork, found the dominant versions by star count. GSD: 48,000 stars. Oh My Claudecode: 25,000 stars. 73,000 combined. Almost nobody in this community talks about either of them. I had both installed before dinner. GSD manages what goes into Claude's context at each phase of your project. Instead of flooding the window with everything at once, it structures exactly what Claude needs, when it needs it. Plans. Specs. Phases. One terminal command installs it as 68 global skills that activate automatically across every project you open. Oh My Claudecode handles multi-agent orchestration. You describe a task. It figures out how to split the work across specialized agents and runs them in parallel. Automatically. Both are free. Both open source. GSD: npx get-shit-done-cc@latest Oh My Claudecode: /plugin marketplace add https://github.com/Yeachan-Heo/oh-my-claudecode Ten minutes. Done. The people pulling the most out of AI right now are not just using the tool. They're upgrading what runs underneath it. You've probably already felt the context rot in your own sessions. You just didn't have a name for it, or a fix. Now you have both. Drop a comment if you want more detail on what each one does in practice.