Interview skipped ahead during teaching
What the user saw
When a candidate asked for an explanation or said they did not know, the interview could use up the current question and move ahead. It could also send voice input after a pause that was too short for thinking.
Why it happened
Both clients advanced the question after every reply, even when the candidate was still learning the current topic.
What the real fix did
It added a teach reply that keeps the same question open on web and mobile, and lengthened the interview voice pause to five seconds.
This issue is worth 2 points: the real fix changed 82 source lines. Models fixed it in 13 of 16 recorded runs.
Seen in
- GPT-6 Luna vs Haiku 5.5: 20 Real Bugs
- Claude Haiku 5.5 vs GPT-6 Luna (Max effort)
- MiniMax M3.1 vs DeepSeek V4.1
- GPT-6.1 Terra vs Claude Sonnet 5.5
- Crouton Alpha vs Tencent Hy4: 20 Real Bugs, One Model Is Free
- Particle-1 vs Kimi K3.1 Pro: 20 Real Bugs
- Claude Fable 5.2 (Medium effort): 20 Real Bugs
- Muse Ember (High effort): 20 Real Bugs
- Muse Spark 1.3 Contributor (High effort): 20 Real Bugs