ChatGPT, Codex Cut Desktop Storage 60%

This is an unglamorous but essential fix for desktop agents that persist increasingly large workflows locally.

• –Addresses a recurring pain point as long-running threads, tool outputs, and subagents accumulate on disk
• –Could significantly extend usable storage for heavy Codex users
• –Community reports have documented local session and runtime directories growing to tens or hundreds of gigabytes
• –The reduction targets stored state, but may not eliminate separate high disk-write or cache-churn issues
• –Better storage management is becoming a core reliability feature for agent workspaces