A recent statistic suggesting widespread user preference for both Codex and Claude Code was likely misinterpreted, according to an analysis of incident reports. The statistic, which indicated that over 60% of T3 Code users who used Codex also used Claude Code, was presented as evidence of interchangeable models and the value of a unified harness. However, an examination of status pages from August 1 to October 19 revealed significant downtime for both services, with Codex experiencing approximately 43 hours of incidents and Claude Code around 34 minutes during a two-week period. When considering an 80-day span, both models had substantial downtime, but they were concurrently unavailable for less than two hours, suggesting that users might be employing both services as a backup rather than due to preference. AI
IMPACT Suggests that users may be employing multiple LLM services as backups due to reliability issues, rather than for feature preference.
RANK_REASON Analysis of public status pages refutes a widely shared statistic about LLM usage, offering an alternative interpretation.
- Anthropic
- Claude Code
- Codex
- Codex API
- Codex in ChatGPT Desktop
- Codex Web
- OpenAI
- T3 Code
- T3.gg
- Theo Browne
- VS Code Extension
- X
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
