DeepSeek's Insane Architecture Breakthrough [Engram Explained]
Unlock all features
FREE: Get instant access to 10 AI summaries, chats, or transcripts per day.
Unlock all features
FREE: Get instant access to 10 AI summaries, chats, or transcripts per day.
Unlock all features
FREE: Get instant access to 10 AI summaries, chats, or transcripts per day.
Unlock all features
FREE: Get instant access to 10 AI summaries, chats, or transcripts per day.
Unlock all features
FREE: Get instant access to 10 AI summaries, chats, or transcripts per day.
Related videos
How DeepSeek V4 Broke AI’s Cost Curse
bycloud
101.8k views
What Is Yann LeCun Cooking? JEPA Explained Simply
bycloud
50.7k views
DeepSeek Just Added Parameters Where There Were None
bycloud
34.5k views
The End of the Steam Age? China’s Breakthrough CO2 Generator
Anton Petrov
86.5k views
The Arena Breaks Down An INSANE Week 16 In The NFL
The Arena
106.6k views
DeepSeek V3.2 Just Broke SoTA Again… But How?
bycloud
168.2k views
Egon Schiele: Great Art Explained
Great Art Explained
161.1k views
WTF: Viral WNBA D*Ido Throwing Explained
Breaking Points
82.7k views
10x Faster Than Standard LLM!? DiffusionLM Explained
bycloud
63.7k views
Elon's Neuralink Breakthrough And Telepathy Is Actually Awesome
Destiny
161.7k views
Top Comments (10)
So we're back to using n-grams. Next up: fine state automata. Qwen 5 will just be one big ass regular expression.
I’ve spent enough time reading model papers to know when something is actually a leap, and this one really does feel like a new building block rather than just another tweak!
Engram? wake the fuck up samurai!
I've never seen such complicated matters explained so well... visually, well done.
its nice that no AI giant has won the competition yet because then winner works only to stay on top. look how that worked out with windows for os landscape.
Btw Deepseek seems to be trying to introduce mHc, tiered kv-cache, engram memory, and optimization for Huawei's GPUs all in one release with V4
Dude never stop making these videos. A real breath of fresh air in the world of AI slop content makers wilding out over every new slop tool. Amazing animations, clear explanations and technical subjects. You are making a big impact making these concepts aproachable.
Paper so old TwoMinutesPapers made a video about this before you
This feels like the origin of "Engram is all you need" arc
Stop scraping manually and check out SerpAPI now! https://serpapi.link/bycloud-march-2026
Unlock the Data Inside
Turn Videos into Knowledge
- Get FREE 10/day: transcripts, summaries, chats
- Chat with videos, export text & PDF
- $1 free API credit for RAG, chatbots & research
Free forever plan • All features unlocked
Top Comments (10)
So we're back to using n-grams. Next up: fine state automata. Qwen 5 will just be one big ass regular expression.
I’ve spent enough time reading model papers to know when something is actually a leap, and this one really does feel like a new building block rather than just another tweak!
Engram? wake the fuck up samurai!
I've never seen such complicated matters explained so well... visually, well done.
its nice that no AI giant has won the competition yet because then winner works only to stay on top. look how that worked out with windows for os landscape.
Btw Deepseek seems to be trying to introduce mHc, tiered kv-cache, engram memory, and optimization for Huawei's GPUs all in one release with V4
Dude never stop making these videos. A real breath of fresh air in the world of AI slop content makers wilding out over every new slop tool. Amazing animations, clear explanations and technical subjects. You are making a big impact making these concepts aproachable.
Paper so old TwoMinutesPapers made a video about this before you
This feels like the origin of "Engram is all you need" arc
Stop scraping manually and check out SerpAPI now! https://serpapi.link/bycloud-march-2026