OpenAI’s Codex is facing significant backlash due to a flawed logging implementation that has resulted in excessive write operations on users’ solid-state drives (SSDs). This issue is not just a technical glitch; it has financial implications for users, as the excessive writes can drastically reduce the lifespan of SSDs.
Excessive Write Operations Identified
A recent bug report highlights the severity of the issue, indicating that Codex’s logging can lead to approximately 640 TB of data written per year. Developer Rui Fan, a project management committee member of Apache Flink, noted that his SSD had written around 37 TB in just 21 days of uptime. This extrapolates to a staggering rate of write operations that could consume the entire write endurance of a typical 1 TB SSD within a year.
Financial Impact on Users
The economic ramifications of this logging issue are significant. One developer reported a loss of $38.64 in drive value due to the excessive writes. Another estimate suggested that this bug could have cost users collectively in the low-single-digit millions of dollars during the period from March to June. This assessment is based on a cost of $0.13 per TB written, calculated from the SSD price and its endurance rating.
OpenAI’s Response and Future Fixes
OpenAI has acknowledged the problem and confirmed that engineers are actively working on a fix. The company stated that the excessive logging was unintended and resulted from high-volume data being stored in a manner that led to increased disk activity. Recent pull requests indicate that the team is making progress, although users continue to report ongoing issues.
Concerns Over Logging Practices
The logging problem appears to stem from the decision to write app-server SQLite logs at a TRACE level, which generates more verbose logs than necessary. This logging was introduced when Codex debuted last year and is enabled by default, meaning users are unaware of the potential impact on their hardware. As OpenAI works to rectify the situation, the incident raises questions about the implications of such logging practices on user hardware and costs.
This article was produced by NeonPulse.today using human and AI-assisted editorial processes, based on publicly available information. Content may be edited for clarity and style.
