DeepSeek, every mention
4 scenes, the whole family · ← back to DeepSeek
tap a year for its mentions
every year anyone Diana Hu 1Ankit Gupta 1Amjad Masad 1
Verbatim, from the transcripts: the passages where DeepSeek comes up
Multi-GPU Kernels, Intelligence per Watt, Heterogeneous Inference, and More | YC Paper Club · Y Combinator
- ▶ 45:47 unnamed speaker Um, similarly, if you want to deploy these kernels in a real inference engine, well, inference engines, like, let's say, out of the box, if you were to use something like SGLang or VLM, if it's just loading DeepSeq before the first…
Anthropic Head of Pretraining on Scaling Laws, Compute, and the Future of AI · Y Combinator
- ▶ 34:09 Ankit Gupta Quen smaller reasoning models distilled off of the larger Quen models, for example, and similar with Deep Seek, for example.
The Future of Software Creation with Replit CEO Amjad Masad · Y Combinator
- ▶ 12:22 Amjad Masad So the hype today's test time compute, if you think about the sort of O-three or like O-series models or DeepSeq R-one, the kind of main insight there is the more tokens the model is able to consume or produce,