Загрузка видео...

Не удалось загрузить видео

На главную

ok so let me explain why subagents kill long context Like you can spend $500m building 100 million context models, and they would be 1) slow, 2) expensive to use, 3) have huge context rot. O(n) is the lower bound. Cog's approach is something you learn in day 1...

swyx's profile picture

swyx 🇸🇬

155,072 subscribers

276,060 просмотров • 8 месяцев назад •via X (Twitter)

Наука и технологииОбразованиеЮмор
LIVE

Anya Rossi• Live Now

Private livecam show

Комментарии: 0

Нет доступных комментариев

Здесь появятся комментарии из оригинального поста

Похожие видео

Introducing Claude Code Hook - Context Timeline (Saving this to try later) Install with: npx claude-code-templates@latest --hook monitoring/context-timeline Managing the context window and the subagents running in Claude Code is hard to keep track of That's why I built this hook... It starts the moment you open a session and shows a timeline with the main agent's context window and how subagents start working in their own separate context Every subagent you have running will show up in real time This way you can manage the context and the subagents you run, and see everything in a much simpler way than in the console
0:40

Introducing Claude Code Hook - Context Timeline (Saving this to try later) Install with: npx claude-code-templates@latest --hook monitoring/context-timeline Managing the context window and the subagents running in Claude Code is hard to keep track of That's why I built this hook... It starts the moment you open a session and shows a timeline with the main agent's context window and how subagents start working in their own separate context Every subagent you have running will show up in real time This way you can manage the context and the subagents you run, and see everything in a much simpler way than in the console

Daniel San

51,642 просмотров • 2 месяцев назад

I cant believe this guy just made a permanent solution to context bloat and open sourced it all! when we tested this tool (Context+) for solving an issue on the OpenCode repository, the agent using this tool used ~6.5k fewer tokens, found the code and fixed it in half the time! the results were surprising: 6 to 10k tokens saved per prompt, completed task in ~2 minutes while the agent running without the tool took ~4 mins for the same and got stuck in loops bro built an entire beast by using all the modern tools that we could think of: undo trees, semantic search by meaning (by haskellforall), advanced refactoring, blast radius, advanced file context trees, restore points... i can keep going on semantic code search and context trees are the future of agentic coding and this tool proves it the feature i loved the most is semantic search and how it gets things done 2x faster with least possible tokens it makes an agent that actually knows what it’s doing and not just guessing, it makes meaning from your code similar to RAG. if you aren't optimizing your context, you are just burning money the developer says this tool is still under development, it can have unexpected behavior and the docs need updates but the video shows the reality of how fast it can be github: get here:
0:46

I cant believe this guy just made a permanent solution to context bloat and open sourced it all! when we tested this tool (Context+) for solving an issue on the OpenCode repository, the agent using this tool used ~6.5k fewer tokens, found the code and fixed it in half the time! the results were surprising: 6 to 10k tokens saved per prompt, completed task in ~2 minutes while the agent running without the tool took ~4 mins for the same and got stuck in loops bro built an entire beast by using all the modern tools that we could think of: undo trees, semantic search by meaning (by haskellforall), advanced refactoring, blast radius, advanced file context trees, restore points... i can keep going on semantic code search and context trees are the future of agentic coding and this tool proves it the feature i loved the most is semantic search and how it gets things done 2x faster with least possible tokens it makes an agent that actually knows what it’s doing and not just guessing, it makes meaning from your code similar to RAG. if you aren't optimizing your context, you are just burning money the developer says this tool is still under development, it can have unexpected behavior and the docs need updates but the video shows the reality of how fast it can be github: get here:

forloop

225,912 просмотров • 4 месяцев назад

RLM is the most import foundation of my Pi Harness (other than Pi of course). It's seeded with late interaction retrieval results (thanks to @lightonai for pylate). The Agent initiates it with query then.. 𝐒𝐞𝐭𝐮𝐩 A python REPL is created and seeded with: 1. Late interaction search to pre-filter. Instead of doing top 3/5/10, it's top hundreds of documents. This is set into a `context` variable. 2. Python functions are loaded in to do more searches if `context` variable isn't enough. And to make llm calls with cheaper models in parallel batches. 𝐈𝐭𝐞𝐫𝐚𝐭𝐢𝐨𝐧 𝐋𝐨𝐨𝐩 From there, an LLM iterates in the REPL based on the query. It's just like exploring in a jupyter notebook. The LLM writes prose (like a markdown cell) and code to be run in the REPL each turn. This allows the LLM to sort, filter, and synthesize information. It can fan out and ask smaller models to summarize, combine, contrast, or do anything else to documents to help it understand the data. After several turns the LLM reponds with the final answer. Either because it found the answer, or hit the budget limit. Context as a Python variable, LLM as the programmer, REPL as the runtime. 𝐖𝐡𝐲 𝐃𝐨𝐞𝐬 𝐓𝐡𝐢𝐬 𝐖𝐨𝐫𝐤 1. Richer Shell. Agents (and subagents) work by intermixing code and prose/thinking. But they use static scripts or bash that run and exit and start over each tool call. That's not ideal for exploration and synthesis of data. For that, state is useful to continue building and exploring the data as you learn more. There's a reason jupyter notebooks have been popular with data scientists. 2. Keeps main agent context clean. The better context you have the better the agent will perform (duh!). This means three thing: better human input, less missing search results, and less incorrect search results. Letting the agent iterate allows it to synthesize just what is needed and nothing else. All bad paths or peeks at something that turns out to be irrelevant stays out of main agent context. 3. Stack the good ideas! People often compare late interaction search vs RLM. Or static vs dynamic languages. Or agentic search vs semantic search. But...You can just use them all together for what they're each good at. Use them all for the area they're really great for. Read the full post which has more detail about how and why.
3:43

RLM is the most import foundation of my Pi Harness (other than Pi of course). It's seeded with late interaction retrieval results (thanks to @lightonai for pylate). The Agent initiates it with query then.. 𝐒𝐞𝐭𝐮𝐩 A python REPL is created and seeded with: 1. Late interaction search to pre-filter. Instead of doing top 3/5/10, it's top hundreds of documents. This is set into a `context` variable. 2. Python functions are loaded in to do more searches if `context` variable isn't enough. And to make llm calls with cheaper models in parallel batches. 𝐈𝐭𝐞𝐫𝐚𝐭𝐢𝐨𝐧 𝐋𝐨𝐨𝐩 From there, an LLM iterates in the REPL based on the query. It's just like exploring in a jupyter notebook. The LLM writes prose (like a markdown cell) and code to be run in the REPL each turn. This allows the LLM to sort, filter, and synthesize information. It can fan out and ask smaller models to summarize, combine, contrast, or do anything else to documents to help it understand the data. After several turns the LLM reponds with the final answer. Either because it found the answer, or hit the budget limit. Context as a Python variable, LLM as the programmer, REPL as the runtime. 𝐖𝐡𝐲 𝐃𝐨𝐞𝐬 𝐓𝐡𝐢𝐬 𝐖𝐨𝐫𝐤 1. Richer Shell. Agents (and subagents) work by intermixing code and prose/thinking. But they use static scripts or bash that run and exit and start over each tool call. That's not ideal for exploration and synthesis of data. For that, state is useful to continue building and exploring the data as you learn more. There's a reason jupyter notebooks have been popular with data scientists. 2. Keeps main agent context clean. The better context you have the better the agent will perform (duh!). This means three thing: better human input, less missing search results, and less incorrect search results. Letting the agent iterate allows it to synthesize just what is needed and nothing else. All bad paths or peeks at something that turns out to be irrelevant stays out of main agent context. 3. Stack the good ideas! People often compare late interaction search vs RLM. Or static vs dynamic languages. Or agentic search vs semantic search. But...You can just use them all together for what they're each good at. Use them all for the area they're really great for. Read the full post which has more detail about how and why.

Isaac Flath

40,212 просмотров • 2 месяцев назад

Subagents have arrived in Gemini CLI! 🤖🚀 Create your own custom subagents in Gemini CLI! Subagents are specialized, expert agents that the main agent can delegate work to. 📦- Subagents have their own set of tools, MCP servers, system instructions, and context window. 🏷️- Use Agent to explicitly delegate to a subagent 🧹- Keeps the main context window clean ⚡️- Speed up work by running agents in parallel Read more in the launch blog below 👇
0:33

Subagents have arrived in Gemini CLI! 🤖🚀 Create your own custom subagents in Gemini CLI! Subagents are specialized, expert agents that the main agent can delegate work to. 📦- Subagents have their own set of tools, MCP servers, system instructions, and context window. 🏷️- Use Agent to explicitly delegate to a subagent 🧹- Keeps the main context window clean ⚡️- Speed up work by running agents in parallel Read more in the launch blog below 👇

Jack Wotherspoon

221,702 просмотров • 2 месяцев назад

In this episode, Beyang and Thorsten discuss strategies for effective agentic coding, including the 101 of how it's different from coding with chat LLMs, the key constraint of the context window, how and where subagents can help, and the new oracle subagent which combines multiple LLMs. 00:38 Intros 03:20 How coding with agents is very different from coding with prior AI tools 10:31 Example: fix a simple issue 14:13 Example: debugging an issue with an MCP server 21:50 Example: unifying two build scripts 25:09 How the context window is a key constraint 31:01 Why it's best to focus on one thing at a time 33:09 Subagents and context windows 33:49 The codebase search subagent 38:33 General-purpose subagents 44:05 When to use subagents 46:44 The oracle subagent and o3 51:32 Multi-model agents
53:54

In this episode, Beyang and Thorsten discuss strategies for effective agentic coding, including the 101 of how it's different from coding with chat LLMs, the key constraint of the context window, how and where subagents can help, and the new oracle subagent which combines multiple LLMs. 00:38 Intros 03:20 How coding with agents is very different from coding with prior AI tools 10:31 Example: fix a simple issue 14:13 Example: debugging an issue with an MCP server 21:50 Example: unifying two build scripts 25:09 How the context window is a key constraint 31:01 Why it's best to focus on one thing at a time 33:09 Subagents and context windows 33:49 The codebase search subagent 38:33 General-purpose subagents 44:05 When to use subagents 46:44 The oracle subagent and o3 51:32 Multi-model agents

Amp — Research Preview

24,534 просмотров • 11 месяцев назад

There is no better place on the internet to learn about Context Engineering than this repo. It's literally a course with a learning path. It gathers the best resources and covers the theory + code for anything related to context, RAG, memory, and agentic systems and more.
0:26

There is no better place on the internet to learn about Context Engineering than this repo. It's literally a course with a learning path. It gathers the best resources and covers the theory + code for anything related to context, RAG, memory, and agentic systems and more.

ℏεsam

48,999 просмотров • 10 месяцев назад

Never bet against Cursor. Subagents solve so many issues between different models it's kind of insane, especially with Composer 1. Composer 1 can quickly gather context for a model like GPT-5.2 Codex, which otherwise could take on the order of 10 mins just to search files. Definitely one of the most underrated models right now. Now you get the benefits of it's speed, but can handoff to a smarter model at implementation. The video I attached is just a simple example, but you can see how long GPT 5.2 takes to gather context on a very simple problem. For harder problems, the search time grows significantly.
0:50

Never bet against Cursor. Subagents solve so many issues between different models it's kind of insane, especially with Composer 1. Composer 1 can quickly gather context for a model like GPT-5.2 Codex, which otherwise could take on the order of 10 mins just to search files. Definitely one of the most underrated models right now. Now you get the benefits of it's speed, but can handoff to a smarter model at implementation. The video I attached is just a simple example, but you can see how long GPT 5.2 takes to gather context on a very simple problem. For harder problems, the search time grows significantly.

ostyn

19,631 просмотров • 5 месяцев назад

Building an Agentic Search System Building an agentic system is not too hard. Loops, function calling, tool execution, and the model. That's it! I show in this video how to build a search agent from scratch. ~350 lines of code!
13:01

Building an Agentic Search System Building an agentic system is not too hard. Loops, function calling, tool execution, and the model. That's it! I show in this video how to build a search agent from scratch. ~350 lines of code!

elvis

57,290 просмотров • 1 год назад

Today, we’re launching the Parallel Search API, the most accurate web search for AI agents, built using our proprietary web index and retrieval infrastructure. Traditional search ranks URLs for humans to click. AI search needs something different: the right tokens in their context window. Parallel’s Search API is built from the ground up for AI agents, resulting in higher accuracy, lower costs, and lower latency for end-to-end agentic workflows.
1:20

Today, we’re launching the Parallel Search API, the most accurate web search for AI agents, built using our proprietary web index and retrieval infrastructure. Traditional search ranks URLs for humans to click. AI search needs something different: the right tokens in their context window. Parallel’s Search API is built from the ground up for AI agents, resulting in higher accuracy, lower costs, and lower latency for end-to-end agentic workflows.

Parallel Web Systems

669,541 просмотров • 8 месяцев назад

GPT-5.4 is here. Native computer-use capabilities. Up to 1M tokens of context in Codex and the API. Best-in-class agentic coding for complex tasks. Scalable tool search across larger ecosystems. More efficient reasoning for long, tool-heavy workflows.
3:02

GPT-5.4 is here. Native computer-use capabilities. Up to 1M tokens of context in Codex and the API. Best-in-class agentic coding for complex tasks. Scalable tool search across larger ecosystems. More efficient reasoning for long, tool-heavy workflows.

OpenAI Developers

1,074,122 просмотров • 4 месяцев назад

How can you solve complex tasks using a Large Language Model? Here is a 2-minute introduction to everything you need to know to 10x the quality of your results. Let's talk about three techniques, in order of complexity, starting with the easiest one: • In-Context Learning • Indexing + In-Context Learning • Fine-tuning In-Context Learning The team that trained GPT-3 found something they couldn't explain: You can condition a model using examples of how you want it to behave. I included an example prompt in the attached video. You can "teach" the model how you want it to interpret questions, select the correct answers, and format the results by giving a few examples. You can also give specific knowledge to the model that will be helpful when formulating answers. We call this approach "grounding the model." There's another example in the video. Indexing + In-Context Learning Unfortunately, there is a limit to how much data you can include in a prompt. We call this the "context size." One version of GPT-4 supports a context of approximately 6,000 words, while the other supports 25,000 words. Although this sounds like a lot, many applications need more than that. Imagine you wrote a book and want to build an application to answer any questions about your story. What happens if your book is longer than the context? That's where Indexing comes in. Using a model, you can turn every book passage into an embedding. These are vectors, numbers that "encode" the passage's text. You can then store these embeddings in a particular database that supports fast retrieval of these vectors. You can then turn any question into an embedding and search the database for the list of passages that are similar to that query. Instead of using the entire book to ask the model, you can now use the relevant passages as in-context information, effectively working around the context size limitation. Fine-tuning Fine-tuning can give you an extra boost to get reliable outputs from your LLM. It is, however, the most complex approach on the list. There are different approaches to fine-tuning a model with your data. A popular technique is to process your data with your LLM and use the outputs to train a new classifier that solves your specific task. Notice that here you aren't modifying the LLM. Instead, you are chaining it with your trained classifier. Another approach is to modify the parameters of the LLM using your data. Think of this as "rewiring" the model in a way that solves your particular task. The results and costs will vary depending on how many layers you want to fine-tune from the original model. Many companies think that fine-tuning is the solution to their problems. In my experience, many will benefit from exploring the other two approaches. I love explaining Machine Learning and Artificial Intelligence ideas. If you enjoy in-depth content like this, follow me Santiago so you don't miss what comes next.
4:50

How can you solve complex tasks using a Large Language Model? Here is a 2-minute introduction to everything you need to know to 10x the quality of your results. Let's talk about three techniques, in order of complexity, starting with the easiest one: • In-Context Learning • Indexing + In-Context Learning • Fine-tuning In-Context Learning The team that trained GPT-3 found something they couldn't explain: You can condition a model using examples of how you want it to behave. I included an example prompt in the attached video. You can "teach" the model how you want it to interpret questions, select the correct answers, and format the results by giving a few examples. You can also give specific knowledge to the model that will be helpful when formulating answers. We call this approach "grounding the model." There's another example in the video. Indexing + In-Context Learning Unfortunately, there is a limit to how much data you can include in a prompt. We call this the "context size." One version of GPT-4 supports a context of approximately 6,000 words, while the other supports 25,000 words. Although this sounds like a lot, many applications need more than that. Imagine you wrote a book and want to build an application to answer any questions about your story. What happens if your book is longer than the context? That's where Indexing comes in. Using a model, you can turn every book passage into an embedding. These are vectors, numbers that "encode" the passage's text. You can then store these embeddings in a particular database that supports fast retrieval of these vectors. You can then turn any question into an embedding and search the database for the list of passages that are similar to that query. Instead of using the entire book to ask the model, you can now use the relevant passages as in-context information, effectively working around the context size limitation. Fine-tuning Fine-tuning can give you an extra boost to get reliable outputs from your LLM. It is, however, the most complex approach on the list. There are different approaches to fine-tuning a model with your data. A popular technique is to process your data with your LLM and use the outputs to train a new classifier that solves your specific task. Notice that here you aren't modifying the LLM. Instead, you are chaining it with your trained classifier. Another approach is to modify the parameters of the LLM using your data. Think of this as "rewiring" the model in a way that solves your particular task. The results and costs will vary depending on how many layers you want to fine-tune from the original model. Many companies think that fine-tuning is the solution to their problems. In my experience, many will benefit from exploring the other two approaches. I love explaining Machine Learning and Artificial Intelligence ideas. If you enjoy in-depth content like this, follow me Santiago so you don't miss what comes next.

Santiago

384,495 просмотров • 3 лет назад

New short course: Build Long-Context AI Apps with Jamba. Learn about state space models (SSMs), which have emerged as an alternative to transformers! Specifically, Jamba is a hybrid transformer-Mamba architecture that combines strengths of the transformer with ideas from SSMs. This course is built with AI21 Labs and taught by Chen Wang and Chen Almagor. The transformer architecture is computationally expensive when handling very long input contexts. But there's an alternative called Mamba, a selective state space model that can process very long contexts with a much lower computational cost. However, researchers found that the pure Mamba architecture underperforms in understanding the context, and gives lower-quality responses. To overcome this, AI21 developed the Jamba model, which combines Mamba's computational efficiency with the transformer's attention mechanism to help with the output quality. In this course, you’ll learn about how state space models, and Jamba, work. You’ll also learn how to prompt Jamba, use it to process long documents, and build long-context RAG apps. - Learn how Jamba combines transformer and state space model architectures to achieve high performance and quality - Use the AI21 SDK, with an example of prompting over a large 200k-token annual financial report of Nvidia - Use Jamba for tool-calling, with hands-on examples from calling simple arithmetic calculations to a function that returns quarterly company financial reports. - Learn how training for long context is done, and the metrics used for its evaluation - Create a RAG app using the AI21 Conversational RAG tool and build your own RAG pipeline that uses Jamba and LangChain. By the end of this course, you'll learn how to build applications that can handle context as long as an entire book. Please sign up here:
2:04

New short course: Build Long-Context AI Apps with Jamba. Learn about state space models (SSMs), which have emerged as an alternative to transformers! Specifically, Jamba is a hybrid transformer-Mamba architecture that combines strengths of the transformer with ideas from SSMs. This course is built with AI21 Labs and taught by Chen Wang and Chen Almagor. The transformer architecture is computationally expensive when handling very long input contexts. But there's an alternative called Mamba, a selective state space model that can process very long contexts with a much lower computational cost. However, researchers found that the pure Mamba architecture underperforms in understanding the context, and gives lower-quality responses. To overcome this, AI21 developed the Jamba model, which combines Mamba's computational efficiency with the transformer's attention mechanism to help with the output quality. In this course, you’ll learn about how state space models, and Jamba, work. You’ll also learn how to prompt Jamba, use it to process long documents, and build long-context RAG apps. - Learn how Jamba combines transformer and state space model architectures to achieve high performance and quality - Use the AI21 SDK, with an example of prompting over a large 200k-token annual financial report of Nvidia - Use Jamba for tool-calling, with hands-on examples from calling simple arithmetic calculations to a function that returns quarterly company financial reports. - Learn how training for long context is done, and the metrics used for its evaluation - Create a RAG app using the AI21 Conversational RAG tool and build your own RAG pipeline that uses Jamba and LangChain. By the end of this course, you'll learn how to build applications that can handle context as long as an entire book. Please sign up here:

Andrew Ng

77,792 просмотров • 1 год назад

Jeff Dean says models hallucinate because their training data is “squishy.” But what’s in the context window is sharp, the exact text or video frame right in front of them. The goal is to bring more of the world into that context. Compression gives it intuition. Context gives it precision.
2:00

Jeff Dean says models hallucinate because their training data is “squishy.” But what’s in the context window is sharp, the exact text or video frame right in front of them. The goal is to bring more of the world into that context. Compression gives it intuition. Context gives it precision.

vitrupo

41,352 просмотров • 2 месяцев назад

What a fun little breakfast chat with swyx Windsurf just released Fast Context. Unlike traditional agentic search, Fast Context searches and retrieves relevant code from your codebase 20x faster. Game changing for developers who want to stay in flow. Watch it in action inside Andrej Karpathy’s nanochat repo. All over a bowl of Swyx's favorite childhood cereal, Raisin Bran. Powered by Cerebras.
2:22

What a fun little breakfast chat with swyx Windsurf just released Fast Context. Unlike traditional agentic search, Fast Context searches and retrieves relevant code from your codebase 20x faster. Game changing for developers who want to stay in flow. Watch it in action inside Andrej Karpathy’s nanochat repo. All over a bowl of Swyx's favorite childhood cereal, Raisin Bran. Powered by Cerebras.

Sarah Chieng

397,034 просмотров • 8 месяцев назад

Sam Altman says the perfect AI is “a very tiny model with superhuman reasoning, 1 trillion tokens of context, and access to every tool you can imagine.” It doesn't need to contain the knowledge - just the ability to think, search, simulate, and solve anything.
0:53

Sam Altman says the perfect AI is “a very tiny model with superhuman reasoning, 1 trillion tokens of context, and access to every tool you can imagine.” It doesn't need to contain the knowledge - just the ability to think, search, simulate, and solve anything.

vitrupo

775,766 просмотров • 1 год назад

Introducing SWE-grep and SWE-grep-mini: Cognition’s model family for fast agentic search at >2,800 TPS. Surface the right files to your coding agent 20x faster. Now rolling out gradually to Windsurf users via the Fast Context subagent – or try it in our new playground!
0:28

Introducing SWE-grep and SWE-grep-mini: Cognition’s model family for fast agentic search at >2,800 TPS. Surface the right files to your coding agent 20x faster. Now rolling out gradually to Windsurf users via the Fast Context subagent – or try it in our new playground!

Cognition

663,797 просмотров • 8 месяцев назад

Slack has rolled out a new plugin that allows you to connect to Slack for search, messaging, document creation and more. Use this to get context from Slack into Claude Code so it has context on what you're working on or to post updates.
0:25

Slack has rolled out a new plugin that allows you to connect to Slack for search, messaging, document creation and more. Use this to get context from Slack into Claude Code so it has context on what you're working on or to post updates.

Thariq

141,579 просмотров • 4 месяцев назад