Sagar Vaze

What Sagar Vaze said

1. MTS · Oct 29, 2026

   Why Closed AI Providers Block Cyber Defense | Sagar Vaze

   6 moments · 15 min

   2:38
   1. 14:05

      Closed-model refusals can leave cyber defenders without recourse

      “these defensive requests in practice look quite similar to the offensive ones.”

   2. 9:52

      Mistral trained its trillion-parameter model on its own compute ProPro pick

      “It's a trillion parameter model trained on 1E25 flops, which is a measure of the amount of compute that went into it.”

   3. 3:26

      Fine-tuning can beat prompting on narrow enterprise tasks

      “doing a small, what we call lower fine tuning, or a small fine tuning of the model on the data which we'd collected for this exact task was way, way better.”

   4. 5:45

      Compute is Mistral's main gap to the leading AI labs

      “we've convinced ourselves that we can do more with less.”

   5. 1:01

      Open weights let firms turn proprietary expertise into a moat

      “imagine you're an engineering company in Germany rather than being worried that the next generation of closed-weights models is going to come and wash over your competitive advantage.”

   6. 16:44

      Mistral trained multimodal capabilities from the start

      “rather than training a model on text and then fine-tuning it on images and text or videos and text from the very beginning it's got images baked into it”