
guru • குரு
1.4K posts

guru • குரு
@gurubavan
building, growing, helping. cofounder @metaplane (acq'd @datadoghq). previously @hubspot @medialab.














No one Twitter feed is as dialed as mine. Literally every other tweet is gold. Muted words = algorithmic control. 812 blocked words 4400 blocked or muted accounts The key is any time you feel outrage, block or mute.

Lots of hot takes on whether it's possible that DeepSeek made training 45x more efficient, but @doodlestein wrote a very clear explanation of how they did it. Once someone breaks it down, it's not hard to understand. Rough summary: * Use 8 bit instead of 32 bit floating point numbers, which gives massive memory savings * Compress the key-value indices which eat up much of the VRAM; they get 93% compression ratios * Do multi-token prediction instead of single-token prediction which effectively doubles inference speed * Mixture of Experts model decomposes a big model into small models that can run on consumer-grade GPUs






We are devastated to share One More Multiverse is shutting down. It was a joy to be part of your adventures. Here is our full statement: multiverse.com/letter

We’re bringing Reforge’s expertise into your favorite tools... 🤯 Reforge can now coach you as you write your PRDs, roadmaps, GTM plans, and more; all without leaving Google Docs, Confluence, Notion, or Coda. Reforge for Chrome is available for members: reforge.com/chrome



