Interconnects

109 Episodes

1. Managing frontier model training organizations (or teams)

   Published: 19/03/2025
2. Gemma 3, OLMo 2 32B, and the growing potential of open-source AI

   Published: 13/03/2025
3. Interviewing Eugene Vinitsky on self-play for self-driving and what else people do with RL

   Published: 12/03/2025
4. Elicitation, the simplest way to understand post-training

   Published: 10/03/2025
5. Where inference-time scaling pushes the market for AI companies

   Published: 5/03/2025
6. GPT-4.5: "Not a frontier model"?

   Published: 28/02/2025
7. Character training: Understanding and crafting a language model's personality

   Published: 26/02/2025
8. Claude 3.7 thonks and what's next for inference-time scaling

   Published: 24/02/2025
9. Grok 3 and an accelerating AI roadmap

   Published: 18/02/2025
10. An unexpected RL Renaissance

    Published: 13/02/2025
11. Deep Research, information vs. insight, and the nature of science

    Published: 12/02/2025
12. Making the U.S. the home for open-source AI

    Published: 5/02/2025
13. Why reasoning models will generalize

    Published: 28/01/2025
14. Interviewing OLMo 2 leads: Open secrets of training language models

    Published: 22/01/2025
15. DeepSeek R1's recipe to replicate o1 and the future of reasoning LMs

    Published: 21/01/2025
16. Let me use my local LMs on Meta Ray-Bans

    Published: 15/01/2025
17. (Voiceover) DeepSeek V3 and the actual cost of training frontier AI models

    Published: 9/01/2025
18. The state of post-training in 2025

    Published: 8/01/2025
19. Quick recap on the state of reasoning

    Published: 2/01/2025
20. (Voiceover) 2024 Interconnects year in review

    Published: 31/12/2024