OpenAI intentionally removed Codex's visible context usage indicator (github.com)

OpenAI's recent update to the Codex Desktop application has removed the visible context and token usage indicator, a feature critical for developers during long coding sessions. Previously, this functionality allowed users to monitor context usage, gauge context-window pressure, and determine the appropriate times to compact, start new threads, or adjust their coding inputs. The absence of this visual cue complicates professional workflows, as developers now lack essential feedback on context health and status. This change is significant for the AI/ML community because it affects the efficiency and effectiveness of coding with Codex, especially for those deeply integrated into professional environments. The feedback from users highlights a regression in the application's UI/UX, emphasizing the need for at least an approximate context health indicator. Suggestions for improvement include a simple status display showing context pressure levels and last compaction status, which could greatly enhance user experience. The removal of this feature opens discussions about the importance of interface design in developing AI tools that cater effectively to user needs, particularly in high-stakes coding scenarios.