ChinaDeepSeek tolerates 2 to 3 months of no output from new hires.
The practitioner credits compute, not taste: Kimi has fewer chips to spare for training novices.
DeepSeek screens for programming-contest medallists, while Kimi asks whether candidates have run large-scale experiments.
The account comes from one anonymous social-media post, and neither lab is quoted.
DeepSeek’s bet is that 6 to 12 months turns a graduate into a full engineer.
How each outlet framed it
- TMTPost
- contrasts DeepSeek's olympiad-medal hiring and 2-3 month ramp with Kimi's hire-for-immediate-combat model, attributes the gap to compute abundance
Sources: TMTPost