A comparative study tested three context compaction strategies—FutureOS, OpenCode, and Codex—on their ability to retain information from agent sessions. FutureOS retained 83% of queryable information, significantly outperforming OpenCode (47%) and Codex (38%), with the key difference being that FutureOS preserves assistant prose while others compress it away. The analysis reveals that tool output dominates context volume but is rarely referenced, while the sparse assistant text is the primary source of follow-up questions.
A technical guide for installing a context compaction tool for Codex that allows the model to automatically compact its context after verified phases. The setup uses an OpenAI-signed runtime with a Python adapter, requires Git and Python 3.9+, and has been tested on macOS with ChatGPT.app. Installation involves cloning a repository, running setup commands, and launching via a Python script.