Thinking Long, but Short: Stable Sequential Test-Time Scaling for Large Reasoning Models | ScienceToStartup | ScienceToStartup