bycloud — transcripts.

1DeepSeek's Deleted Vision Paper Is Nuts... 2LLM that loops instead of Doing Chain-of-Thought 3How Did A Chinese Phone Company Topped Open Source LLM? 4The Insane Infrastructure Design of DeepSeek V4 5How Did DeepSeek Make V4 So Cheap? 6An Insanely Elegant LLM Architecture Breakthrough Just Dropped 7AI Simulated OS Is Absurd 8A new way to fine-tune LLMs just dropped 9What Is Yann LeCun Cooking? JEPA Explained Simply 10"Claude Mythos Found Thousands of Zero-days..." 11Google's TurboQuant Memory Reduction Claim vs Reality 12This Simple Trick Made ALL LLMs 2x Faster 13Why can’t LLMs just LEARN the context window? 14DeepSeek's Insane Architecture Breakthrough [Engram Explained] 15The Death of RAG? Recursive LM Explained 16The Implosion of the Top Open Source Lab Qwen 17Kimi K2.5 Brought Us 3 brand NEW LLM Frontier!? 18Need more coding agents? just spawn them on the cloud w/ Oz by Warp 19DeepSeek Just Added Parameters Where There Were None 20The AI Bubble Will Be Impossible To Pop Thanks To... AI Video?! 21LLM’s Billion Dollar Problem 22How Chinese DoorDash Is Making Better LLMs Than Meta 23The RL Irony in LLMs (and its insane new meta) 24The New AI Open Source Trifecta 25DeepSeek V3.2 Just Broke SoTA Again… But How? 26How did a 27M Model even beat ChatGPT? 27The biggest Mystery of LLMs have just been solved 28The Chinese AI Iceberg 29Researchers Are Getting Really Creative Training LLMs [Token Order Prediction] 30The Art of Serving LLMs Efficiently: Context Engineering Explained 31Is It EVEN Possible To Reverse Engineer AI’s Training Data? 32New AI Paradigm?! Energy-Based Transformers Explained 33New AI Meta: Train LLMs To Explore On "Hard" Tokens [RLVR + Entropy] 34Kimi K2 Technical Breakdown: How It Challenged AI’s 7-Year Status Quo 35a rather unusual face reveal 36If AI's So Good, Why Can't It Make Its Own Data? [SEAL] 3710x Faster Than Standard LLM!? DiffusionLM Explained 38The Side Effects of Overusing ChatGPT For Homework 39POV: Chinese AI Lab Teaching Everyone How To Save Millions of Dollars 40What If We Remove Tokenization In LLMs? 41When AI Is Designed Like A Biological Brain 42The LLM's RL Revelation We Didn't See Coming 431-Bit LLM: The Most Efficient LLM Possible? 44The REAL AI Architecture That Unifies Vision & Language 45How DeepSeek Built The Current "Best" Math Prover AI 46AI News: DeepSeek-R1 V2, the new open source SoTA! 47Why would anyone let LLMs predict 4 tokens at once? Multi-Token Prediction Explained 48The Biggest "Lie" in AI? LLM doesn't think step-by-step 49Attention Sink: The Fluke That Made LLMs Actually Usable 50Anthropic Just Proved Reasoning AIs Would Silently Cheat