Loading video...

Video Failed to Load

There was a problem loading this video. This could be due to a temporary network issue or the video might be unavailable.

Claude 3.7 reasoning and coding capabilities are no joke! Watch how I use it to one-shot a quick simulator of how attention mechanisms work. I think it could be awesome for each of us to have access to a personal Andrej Karpathy-like tutor to explain complex stuff to us. 😀

elvis

309,783 subscribers

31,669 views • 1 year ago •via X (Twitter)

Education Science & Technology Arts

Anya Rossi• Live Now

Private livecam show

8 Comments

elvis1 year ago

This is not perfect by any means and I didn't check for mistakes but just wanted to quickly share this to inspire more of these use cases.

elvis1 year ago

I also like this one (Precision & Recall). Claude 3.7 Sonnet (with extended thinking) also did this oneshot. The quality of the code and dashboards is on another level.

elvis1 year ago

Prompts for first example: "Can you help me explain attention mechanisms in Transformers to college students? Think deeply about clever ways to explain the concepts without focusing too much on maths." "Now create a simulator that could students understand it better."

Jay Rodge1 year ago

@karpathy This is a cool use case! Can you share the prompt in this thread?

elvis1 year ago

@karpathy Sure, let me do that.

Tariq Gadgetman1 year ago

@karpathy Claude 3.7's capabilities are indeed impressive! A tutor like Karpathy would make complex concepts much easier to grasp.

Jilong | We provide AI marketer - 24/7 marketing1 year ago

@karpathy personal tutors could boost learning, for sure.

Yang1 year ago

Want to learn how practical AI skills and automations for your business and work? Check out our step-by-step video tutorials 100% FREE 🥳

Related Videos

Claude Code with Sonnet 4.5 is actually incredible I gave it a prompt for a super complex app, and it one shot the entire thing In this video I walk you through how to use Claude Code to build a prompt library app you can start using immediately (no coding experience required):

Claude Code with Sonnet 4.5 is actually incredible I gave it a prompt for a super complex app, and it one shot the entire thing In this video I walk you through how to use Claude Code to build a prompt library app you can start using immediately (no coding experience required):

Alex Finn

40,955 views • 9 months ago

Excited to present the LLM-Council skill. Initial idea by Karpathy. I just packaged it as a skill. You can easily spin up a council of LLMs or agents via Fireworks AI. Watch how the new GLM-5 model "deliberates" on other LLMs' thoughts on the big question, "Can LLMs reason?" Things worth paying attention to: New open models like GLM-5 have surprisingly improved on complex reasoning and long-running agentic tasks. The AskUserQuestion tool in Claude Code came in handy to select the council and chairperson. As Andrej Karpathy puts it, it's a really interesting way to get different perspectives from LLMs, which can lead to better decision-making on whatever task you are working on. You can use it for other agentic coding use cases, like evaluation, tool building, designing, and research.

Excited to present the LLM-Council skill. Initial idea by Karpathy. I just packaged it as a skill. You can easily spin up a council of LLMs or agents via Fireworks AI. Watch how the new GLM-5 model "deliberates" on other LLMs' thoughts on the big question, "Can LLMs reason?" Things worth paying attention to: New open models like GLM-5 have surprisingly improved on complex reasoning and long-running agentic tasks. The AskUserQuestion tool in Claude Code came in handy to select the council and chairperson. As Andrej Karpathy puts it, it's a really interesting way to get different perspectives from LLMs, which can lead to better decision-making on whatever task you are working on. You can use it for other agentic coding use cases, like evaluation, tool building, designing, and research.

elvis

39,452 views • 5 months ago

"Explain to me how the rest of us have houses that have 3,4 toilets, there is a citizen in Kenya who doesn't have a toilet and has to pay to use a toilet, and it is supposed to be okay? How is that okay?"~Ruto

"Explain to me how the rest of us have houses that have 3,4 toilets, there is a citizen in Kenya who doesn't have a toilet and has to pay to use a toilet, and it is supposed to be okay? How is that okay?"~Ruto

Lightcast TV Kenya

51,833 views • 1 month ago

Andrej Karpathy builds a personal wiki to think with LLMs. I took that idea and went further. Claude Code + Obsidian = an AI that actually knows you. Your goals. Your context. Your history. Not a chatbot. A second brain that remembers. Here's exactly how to build it:

Andrej Karpathy builds a personal wiki to think with LLMs. I took that idea and went further. Claude Code + Obsidian = an AI that actually knows you. Your goals. Your context. Your history. Not a chatbot. A second brain that remembers. Here's exactly how to build it:

leopardracer

221,465 views • 3 months ago

Bolt's agent now runs Claude 3.7 Sonnet—a huge upgrade for coding, design & debugging! But that’s not all… We’re introducing Dynamic Reasoning, a first-of-its-kind feature that lets the AI choose how hard to “think” when tackling complex problems.

Bolt's agent now runs Claude 3.7 Sonnet—a huge upgrade for coding, design & debugging! But that’s not all… We’re introducing Dynamic Reasoning, a first-of-its-kind feature that lets the AI choose how hard to “think” when tackling complex problems.

bolt.new

70,684 views • 1 year ago

i don’t know how to explain it but bollywood doesn't give us nepo kids like it used to

i don’t know how to explain it but bollywood doesn't give us nepo kids like it used to

saif

36,561 views • 2 months ago

I found an excellent use case for Remotion & claude code - get it to create a video explaining how the codebase works. In this case I got it to explain how codebase works

I found an excellent use case for Remotion & claude code - get it to create a video explaining how the codebase works. In this case I got it to explain how codebase works

Peter Gostev

15,379 views • 5 months ago

Here’s how to build a Slack clone with Claude 3.7 Sonnet and Cursor Agent in v0.46. Perfect for *anyone* wanting to pick up AI coding tools. Learn how to build an app 100% from scratch with the newest AI coding workflows. Watch for 1hr+ of agentic coding.

Here’s how to build a Slack clone with Claude 3.7 Sonnet and Cursor Agent in v0.46. Perfect for anyone wanting to pick up AI coding tools. Learn how to build an app 100% from scratch with the newest AI coding workflows. Watch for 1hr+ of agentic coding.

Mckay Wrigley

212,434 views • 1 year ago

🐹 I’m sure some of us have regrets, but for me, I think it’s enough that we’ve been to the top once. I think it would be a good idea if each of us lives doing what we’d like to do as we move forward.

🐹 I’m sure some of us have regrets, but for me, I think it’s enough that we’ve been to the top once. I think it would be a good idea if each of us lives doing what we’d like to do as we move forward.

AFJ

25,086 views • 2 years ago

#F1Testing | Lewis Hamilton on what’s on the checklist for the rest of the week: “Hoping to experience the car in the dry, and I think understanding balance, understanding the new SM that we have and how to utilise it. I’ve not driven in the dry so I don’t have a feel for it.” “Understanding the deployment, I think deployment is gonna be crucial and understanding how to utilise the power, the battery throughout the lap and recharging…” “We’ve done a huge amount of work,everyone’s done a huge amount of work in the simulator so it’s not like it’s 100% brand new for us, but we still have to go through our sessions and the data to optimise it.”

deni

23,416 views • 5 months ago

How to build a standout brand in 2026 ✨ I’m working on a YouTube video where I show my full process of building a brand identity from scratch - how I use Midjourney to create an awesome style, what tools I use, and the coolest part, how to build your own tools with AI and vibe coding to create unique textures, effects, and animations no one else has. If this sounds like something you’d watch, share it and drop a kind word below. 👇

Alex Socoloff

71,333 views • 3 months ago

some of you ask me about stuff like how do I create collar/similar stuff like that. Here is a short video for you all, I usually start with a sphere and zremesh it afterward to have clean topology to work on. Hope it is helpful to you all.

some of you ask me about stuff like how do I create collar/similar stuff like that. Here is a short video for you all, I usually start with a sphere and zremesh it afterward to have clean topology to work on. Hope it is helpful to you all.

weilunart

64,013 views • 2 years ago

Is RFK Jr compromised or lying to us? If you want to know how I really feel, take the time to watch this video. Too many of my words are being misinterpreted and I think hearing me explain it will help.

Is RFK Jr compromised or lying to us? If you want to know how I really feel, take the time to watch this video. Too many of my words are being misinterpreted and I think hearing me explain it will help.

Anna Matson

14,535 views • 1 year ago

I wrote tidycensus for #rstats to simplify the process of acquiring US Census data. But - you still have to know R to use it! I wrote a rudimentary tidycensus #MCP to teach Claude to fetch Census data - watch the video for the results.

I wrote tidycensus for #rstats to simplify the process of acquiring US Census data. But - you still have to know R to use it! I wrote a rudimentary tidycensus #MCP to teach Claude to fetch Census data - watch the video for the results.

Kyle Walker

10,855 views • 1 year ago

I built a Claude skill that helps you review a report as an analyst. Now, it thinks like a senior analyst for you. The readme file will explain how to use it as a skill on Claude or Copilot or Chatgpt Use & let me how it works. My gift to U

I built a Claude skill that helps you review a report as an analyst. Now, it thinks like a senior analyst for you. The readme file will explain how to use it as a skill on Claude or Copilot or Chatgpt Use & let me how it works. My gift to U

David Abu _

38,839 views • 4 months ago

OpenAI just announced API access to o1 (advanced reasoning model) yesterday. I'm delighted to announce today a new short course, Reasoning with o1, built with OpenAI, and taught by Colin Jarvis, Head of AI Solutions at OpenAI, to show you how to use this effectively! Unlike previous language models which generate output directly, o1 “thinks before it responds,” and generates many reasoning tokens before returning a more thoughtful and accurate response. It is great at complex reasoning -- including planning for agentic workflows, coding, and domain-specific reasoning in STEM fields like law. But how you should use it is quite different from other LLMs. I think o1 will be a game changer for many AI applications; and in this course, you'll learn how to use it effectively. In detail, you’ll: - Learn to recognize what tasks o1 is suited for, and when to use a smaller model, or combine o1 with a smaller model - Understand the new principles of prompting reasoning models: Be simple and direct; no explicit chain-of-thought required; use structure; show rather than tell - Implement multi-step orchestration in which o1 plans, and hands tasks over to gpt-4o-mini to execute specific steps; this illustrates a design pattern to optimize intelligence (accuracy) and cost - Use o1 for a coding task to build a new application, edit existing code, and test performance by running a coding competition between o1-mini and GPT 4o - Use o1 for image understanding and learn how it performs better with a "hierarchy of reasoning," in which it incurs the latency and cost upfront, preprocessing the image and indexing it with rich details so it can be used for Q&A later - Learn a technique called meta-prompting, in which you use o1 to improve your prompts. Using a customer support evaluation set, you'll iteratively use o1 to modify a prompt to improve performance You'll also learn about how OpenAI used reinforcement learning to produce a model that uses "test-time compute" to improve performance. I think you'll find this course enjoyable and valuable. Please sign up for it here:

OpenAI just announced API access to o1 (advanced reasoning model) yesterday. I'm delighted to announce today a new short course, Reasoning with o1, built with OpenAI, and taught by Colin Jarvis, Head of AI Solutions at OpenAI, to show you how to use this effectively! Unlike previous language models which generate output directly, o1 “thinks before it responds,” and generates many reasoning tokens before returning a more thoughtful and accurate response. It is great at complex reasoning -- including planning for agentic workflows, coding, and domain-specific reasoning in STEM fields like law. But how you should use it is quite different from other LLMs. I think o1 will be a game changer for many AI applications; and in this course, you'll learn how to use it effectively. In detail, you’ll: - Learn to recognize what tasks o1 is suited for, and when to use a smaller model, or combine o1 with a smaller model - Understand the new principles of prompting reasoning models: Be simple and direct; no explicit chain-of-thought required; use structure; show rather than tell - Implement multi-step orchestration in which o1 plans, and hands tasks over to gpt-4o-mini to execute specific steps; this illustrates a design pattern to optimize intelligence (accuracy) and cost - Use o1 for a coding task to build a new application, edit existing code, and test performance by running a coding competition between o1-mini and GPT 4o - Use o1 for image understanding and learn how it performs better with a "hierarchy of reasoning," in which it incurs the latency and cost upfront, preprocessing the image and indexing it with rich details so it can be used for Q&A later - Learn a technique called meta-prompting, in which you use o1 to improve your prompts. Using a customer support evaluation set, you'll iteratively use o1 to modify a prompt to improve performance You'll also learn about how OpenAI used reinforcement learning to produce a model that uses "test-time compute" to improve performance. I think you'll find this course enjoyable and valuable. Please sign up for it here:

Andrew Ng

357,661 views • 1 year ago

Lamar Jackson on Jesse Minter and his Ravens coaching staff: "I feel like our coaches are like one of us, in a sense. ... How he coaches, how on point he is with the schematics of the defense. It's like bro, it's awesome to have. Awesome to have this feeling right now." via Baltimore Ravens

Lamar Jackson on Jesse Minter and his Ravens coaching staff: "I feel like our coaches are like one of us, in a sense. ... How he coaches, how on point he is with the schematics of the defense. It's like bro, it's awesome to have. Awesome to have this feeling right now." via Baltimore Ravens

Bobby Trosset

51,201 views • 1 month ago

I think people underestimate how rare it is for a franchise to maintain The Last of Us level of storytelling consistency.

I think people underestimate how rare it is for a franchise to maintain The Last of Us level of storytelling consistency.

Nyx

10,853 views • 2 months ago

Adriana Chechik revealed that her public s3x fetish wasn’t just about making great art or creating memorable scenes. She explains that the truest part of it was seeing what her body could handle, pushing herself to the limit, and finding out how long she could last. Adriana says she found it powerful to be in a room full of men and still be the one ready to keep going, insisting that she was always the one in control of the experience and the vision behind it. “I like to push myself I like to see like how much d1ck can I take how long can I last and stuff like that and I think it's very powerful to take down a room of men or a group of people and then be the one that's still like I'm ready to go”

Adriana Chechik revealed that her public s3x fetish wasn’t just about making great art or creating memorable scenes. She explains that the truest part of it was seeing what her body could handle, pushing herself to the limit, and finding out how long she could last. Adriana says she found it powerful to be in a room full of men and still be the one ready to keep going, insisting that she was always the one in control of the experience and the vision behind it. “I like to push myself I like to see like how much d1ck can I take how long can I last and stuff like that and I think it's very powerful to take down a room of men or a group of people and then be the one that's still like I'm ready to go”

Fazzy

70,149 views • 1 month ago