The Death of RAG?
Related videos
The unexpected death of Codex
Theo - t3․gg
42.2k views
LLM that loops instead of Doing Chain-of-Thought
bycloud
29.0k views
Ben Shapiro Reacts To The Death Of Spirit Airlines
Ben Shapiro
16.3k views
The RL Irony in LLMs
bycloud
23.0k views
THE DEATH OF MAGA | The Kyle Kulinski Show
Secular Talk
154.0k views
The biggest Mystery of LLMs have just been solved
bycloud
102.6k views
The Chinese AI Iceberg
bycloud
107.4k views
The Death of Todd Stermer | Full Episode
48 Hours
1.1m views
The Death of Streaming
penguinz0
1.7m views
The Side Effects of Overusing ChatGPT For Homework
bycloud
26.6k views
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
The unexpected death of Codex
Theo - t3․gg
42.2k views
LLM that loops instead of Doing Chain-of-Thought
bycloud
29.0k views
Ben Shapiro Reacts To The Death Of Spirit Airlines
Ben Shapiro
16.3k views
The RL Irony in LLMs
bycloud
23.0k views
THE DEATH OF MAGA | The Kyle Kulinski Show
Secular Talk
154.0k views
The biggest Mystery of LLMs have just been solved
bycloud
102.6k views
The Chinese AI Iceberg
bycloud
107.4k views
The Death of Todd Stermer | Full Episode
48 Hours
1.1m views
The Death of Streaming
penguinz0
1.7m views
The Side Effects of Overusing ChatGPT For Homework
bycloud
26.6k views
Top Comments (10)
It's basically the same idea used in Claude Code or in Codex with subagents.
i remember reading infini-attention paper too i believe it when i see it working....
this isn't really a recursive language model -- all of the other papers on this topic talk about recursion happening in the model architecture. This is just a context orchestration algorithm, like a tool chain or agent loop. Not an LM architecture.
The needle in a heystack test should start utilizing vagueness in the query more.
9:00 and you just described RLM.... I really don't think why we need a distinction between RLM and an Agent, I mean in a sense RLM is agentic context management....
**Peeks inside**: Its just rag with more LLMs with subagent pattern. Also RAG is not a vector database. Rag is retrieval augmented generation. It doesnt matter where or how you are retrieving it. RAG is an umbrella term for an LLM using retrieval via any method or tools.
good idea . ill have my agents work on it
Check out Inngest and let your AI agents wear a harness now! https://www.inngest.com/docs?utm_source=youtube&utm_medium=video&utm_campaign=yt-bycl-3
This is already how a lot of us uses agents in for example OpenCode, one root orchestrator, which has a high level overview, gives direct commands to sub-agents, which does context heavy work, and then return compact data to the orchistrator. In my case, the orchstrator, after a 4-5 hour run, has only reached like 150k tokens, which is great. The overall token usage at this point is over 10m.
Interesting! I had been tinkering around with trying to build something similar, as I've always had the belief the models would perform much better in a divide-and-conquer strategy rather than trying to do everything all at once. Glad to see this is also actively being researched
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)
It's basically the same idea used in Claude Code or in Codex with subagents.
i remember reading infini-attention paper too i believe it when i see it working....
this isn't really a recursive language model -- all of the other papers on this topic talk about recursion happening in the model architecture. This is just a context orchestration algorithm, like a tool chain or agent loop. Not an LM architecture.
The needle in a heystack test should start utilizing vagueness in the query more.
9:00 and you just described RLM.... I really don't think why we need a distinction between RLM and an Agent, I mean in a sense RLM is agentic context management....
**Peeks inside**: Its just rag with more LLMs with subagent pattern. Also RAG is not a vector database. Rag is retrieval augmented generation. It doesnt matter where or how you are retrieving it. RAG is an umbrella term for an LLM using retrieval via any method or tools.
good idea . ill have my agents work on it
Check out Inngest and let your AI agents wear a harness now! https://www.inngest.com/docs?utm_source=youtube&utm_medium=video&utm_campaign=yt-bycl-3
This is already how a lot of us uses agents in for example OpenCode, one root orchestrator, which has a high level overview, gives direct commands to sub-agents, which does context heavy work, and then return compact data to the orchistrator. In my case, the orchstrator, after a 4-5 hour run, has only reached like 150k tokens, which is great. The overall token usage at this point is over 10m.
Interesting! I had been tinkering around with trying to build something similar, as I've always had the belief the models would perform much better in a divide-and-conquer strategy rather than trying to do everything all at once. Glad to see this is also actively being researched