正在加载视频...

视频加载失败

加载此视频时出现问题。这可能是由于临时网络问题，或视频可能不可用。

Releasing this tomorrow: A framework for recreating Grok Heavy functionality. In this video, I'm using Claude 4 Heavy, but it works with any model (GPT, Kimi, Gemini).

Pietro Schirano

85,916 subscribers

62,603 次观看 • 1 年前 •via X (Twitter)

科学技术教育艺术

Anya Rossi• Live Now

Private livecam show

11 条评论

daniel inderos 的头像

daniel inderos1 年前

How much does it improve the model’s performance?

Mobile Scanner 的头像

Mobile Scanner1 年前

Scan any documents, convert images into text, PDF files, etc. 👍

Matt Shumer 的头像

Matt Shumer1 年前

Nice work! This will be very cool.

Pietro Schirano 的头像

Pietro Schirano1 年前

Thank you!

Misha Laskin 的头像

Misha Laskin1 年前

Very cool!

🐧 lalo adrian morales 𝕏 的头像

🐧 lalo adrian morales 𝕏1 年前

I dont know anyone else that creates the kind of stuff you do. much respect

Alvaro Cintas 的头像

Alvaro Cintas1 年前

This looks awesome!

•UtilityCo 的头像

•UtilityCo1 年前

Very cool. Does your framework handle the orchestration layer or mostly prompt logic? We’re hacking on decentralized agent swarms for factory automation and could use a portable Grok-style brain on our edge GPUs. Would love to hear how it scales and if you’re open-sourcing the hooks.

🍓🍓🍓 的头像

🍓🍓🍓1 年前

i think something like this with multiple models using some variety of 'deepthink' would be very impressive. you'd need some magical router that knows when to pick the perfect model for the task (or models)

Shariq Riaz 的头像

Shariq Riaz1 年前

This is fucking amazing i was waiting for someone to do this!!!

The Canaanite 的头像

The Canaanite1 年前

Wow

相关视频

Introducing "Make it Heavy" a framework to simulate Grok Heavy functionality, fully open-source and running in your terminal. Use any model you want, thanks to . 👇

Introducing "Make it Heavy" a framework to simulate Grok Heavy functionality, fully open-source and running in your terminal. Use any model you want, thanks to . 👇

Pietro Schirano

181,222 次观看 • 1 年前

For the remaining skeptics who somehow don’t trust the *five* Grok share links above, here’s a full 5 minute video of Grok 4 Heavy answering “Hitler”—starting with a view of my custom instruction settings to show I’m not using any. (And, yes, Grok 4 Heavy really is this slow.)

For the remaining skeptics who somehow don’t trust the five Grok share links above, here’s a full 5 minute video of Grok 4 Heavy answering “Hitler”—starting with a view of my custom instruction settings to show I’m not using any. (And, yes, Grok 4 Heavy really is this slow.)

Riley Goodside

16,814 次观看 • 1 年前

Its here ! Grok 4.5 is now the model being used for Expert and Heavy modes on Grok Web and Apps. Using Grok Heavy now shows 4 agents using Grok v9 aka Grok 4.5 where it previously used 16 agents with the older models. Now we can use SpaceXAI's latest 1.5T parameter model anywhere!

Its here ! Grok 4.5 is now the model being used for Expert and Heavy modes on Grok Web and Apps. Using Grok Heavy now shows 4 agents using Grok v9 aka Grok 4.5 where it previously used 16 agents with the older models. Now we can use SpaceXAI's latest 1.5T parameter model anywhere!

️

68,272 次观看 • 10 天前

Fourier Series: The Math Behind Modern Technology. This entire video was created using Grok 4 Heavy.

Fourier Series: The Math Behind Modern Technology. This entire video was created using Grok 4 Heavy.

tetsuo

46,608 次观看 • 1 年前

Grok 4 Heavy is better than any model available at identifying issues in your codebase. Here's the JS prompt I use with my game code to have Grok 4 Heavy find the bugs. Python prompt in Comments👇

Grok 4 Heavy is better than any model available at identifying issues in your codebase. Here's the JS prompt I use with my game code to have Grok 4 Heavy find the bugs. Python prompt in Comments👇

tetsuo

1,814,559 次观看 • 1 年前

🚨 Gemini 3.0 Heavy same in style like grok heavy , 4 agents work together in paraller , reflect and synthesised in one output One shotted the planet visualization and to access it check comment for direct link and share your results and feedback

🚨 Gemini 3.0 Heavy same in style like grok heavy , 4 agents work together in paraller , reflect and synthesised in one output One shotted the planet visualization and to access it check comment for direct link and share your results and feedback

Chetaslua

85,235 次观看 • 8 个月前

‼️Me and my PIE versus five AIs: Grok, Chat GPT, Claude 3, Gemini, and Google AI. 🧠This video is shows me refuting AI pushback against my epistemic model, proving that it does indeed escape the Agrippa Trilemma, and that this 2000-year-old skeptic trap is a false trichotomy.

‼️Me and my PIE versus five AIs: Grok, Chat GPT, Claude 3, Gemini, and Google AI. 🧠This video is shows me refuting AI pushback against my epistemic model, proving that it does indeed escape the Agrippa Trilemma, and that this 2000-year-old skeptic trap is a false trichotomy.

Lucas Gage

26,871 次观看 • 6 个月前

SwiftUI friends… Any ideas on recreating this? Any built-in functionality I’m not aware of? Trying to create a cool new interaction for Stuff | ToDos and Lists

SwiftUI friends… Any ideas on recreating this? Any built-in functionality I’m not aware of? Trying to create a cool new interaction for Stuff | ToDos and Lists

austin blake (stuff todos+lists)

110,769 次观看 • 9 个月前

Got an output from the strongest pre-release Gemini 3.5 Pro checkpoint 👀 I compared it to Claude Fable 5 and GPT-5.5 on the same exact prompt Yeah... This model is mostly dead on arrival (tried it on other prompts too ofc). It will likely be the "poor man's model" next to 3.5 flash. 1st video: Gemini 3.5 Pro (pre-release) 2nd video: GPT-5.5 3rd video: Claude Fable 5

Got an output from the strongest pre-release Gemini 3.5 Pro checkpoint 👀 I compared it to Claude Fable 5 and GPT-5.5 on the same exact prompt Yeah... This model is mostly dead on arrival (tried it on other prompts too ofc). It will likely be the "poor man's model" next to 3.5 flash. 1st video: Gemini 3.5 Pro (pre-release) 2nd video: GPT-5.5 3rd video: Claude Fable 5

Lentils

50,395 次观看 • 1 个月前

Introducing Codespace A virtual computer that runs custom Claude Code on server. It is using 2 models: Kimi K2 + Claude Sonnet 4 in a sync. I achieved same code quality with 2x better reasoning and 53% less model cost. Access it today CodeGuide (500 credits for everyone)

Introducing Codespace A virtual computer that runs custom Claude Code on server. It is using 2 models: Kimi K2 + Claude Sonnet 4 in a sync. I achieved same code quality with 2x better reasoning and 53% less model cost. Access it today CodeGuide (500 credits for everyone)

CJ Zafir

31,070 次观看 • 11 个月前

We’ve been building this in secret for a while. We love Claude Code, and we’ve reimplemented “Claude Code” in Genspark. Introducing Genspark AI Developer - “Claude Code” in the browser - “Claude Code” on mobile - Everyone can build, no coding experience required - Supports multiple frontier models: Claude Sonnet 4, Claude Opus 4.1, GPT‑5, Kimi + Groq 1/4

We’ve been building this in secret for a while. We love Claude Code, and we’ve reimplemented “Claude Code” in Genspark. Introducing Genspark AI Developer - “Claude Code” in the browser - “Claude Code” on mobile - Everyone can build, no coding experience required - Supports multiple frontier models: Claude Sonnet 4, Claude Opus 4.1, GPT‑5, Kimi + Groq 1/4

Eric Jing

23,245,863 次观看 • 11 个月前

As promised, I built a small local RAG agent to demonstrate how Model Context Protocol (𝐌𝐂𝐏) works in practice when connecting LLMs to tools and data. The agent can read, search, and list my notes; but the functionality itself isn’t the point, the protocol is. Instead of writing Gemini specific tool integrations, I connected the model to an 𝐌𝐂𝐏 server. This is why 𝐌𝐂𝐏 is highly relevant for production AI systems: • 𝐒𝐭𝐚𝐧𝐝𝐚𝐫𝐝𝐢𝐳𝐚𝐭𝐢𝐨𝐧: one protocol instead of custom connectors per data source. • 𝐏𝐨𝐫𝐭𝐚𝐛𝐢𝐥𝐢𝐭𝐲: the same MCP server works with Gemini, Claude, or any MCP compliant client. • 𝐒𝐚𝐟𝐞𝐭𝐲: models can only access explicitly exposed capabilities. 𝐌𝐂𝐏 shifts agent development from integration heavy to capability driven. Demo link in the first comment.

As promised, I built a small local RAG agent to demonstrate how Model Context Protocol (𝐌𝐂𝐏) works in practice when connecting LLMs to tools and data. The agent can read, search, and list my notes; but the functionality itself isn’t the point, the protocol is. Instead of writing Gemini specific tool integrations, I connected the model to an 𝐌𝐂𝐏 server. This is why 𝐌𝐂𝐏 is highly relevant for production AI systems: • 𝐒𝐭𝐚𝐧𝐝𝐚𝐫𝐝𝐢𝐳𝐚𝐭𝐢𝐨𝐧: one protocol instead of custom connectors per data source. • 𝐏𝐨𝐫𝐭𝐚𝐛𝐢𝐥𝐢𝐭𝐲: the same MCP server works with Gemini, Claude, or any MCP compliant client. • 𝐒𝐚𝐟𝐞𝐭𝐲: models can only access explicitly exposed capabilities. 𝐌𝐂𝐏 shifts agent development from integration heavy to capability driven. Demo link in the first comment.

Sui_Builds

12,853 次观看 • 6 个月前

Want to vibe code games for iOS easily? Steal this workflow. > Love2D engine > 2 agent skills > Write once; works everywhere (mac + simulator + device) > No need to open Xcode I used Claude Code Kimi K2.5, but works with Opus 4.5 / GPT 5.2, etc. FULL step-by-step video 👇

Want to vibe code games for iOS easily? Steal this workflow. > Love2D engine > 2 agent skills > Write once; works everywhere (mac + simulator + device) > No need to open Xcode I used Claude Code Kimi K2.5, but works with Opus 4.5 / GPT 5.2, etc. FULL step-by-step video 👇

Chong-U

16,929 次观看 • 5 个月前

This video game doesn't utilise any ray tracing, but still looks absolutely stunning! I can't believe it runs this buttery smooth with such heavy foliage at max graphics.

This video game doesn't utilise any ray tracing, but still looks absolutely stunning! I can't believe it runs this buttery smooth with such heavy foliage at max graphics.

John

42,678 次观看 • 22 天前

Someone built a free and better alternative to Claude that runs 100% locally. → works with any LLM (Claude, GPT, Gemini, vLLM) → beats it on deep research → has Cowork-like capabilities → 50+ connectors out of the box → deploy in literally one command 100% open source. MIT license. 28k stars.

Someone built a free and better alternative to Claude that runs 100% locally. → works with any LLM (Claude, GPT, Gemini, vLLM) → beats it on deep research → has Cowork-like capabilities → 50+ connectors out of the box → deploy in literally one command 100% open source. MIT license. 28k stars.

How To Prompt

39,202 次观看 • 2 个月前

Coded with Grok 4 from xAI , built this Three.js demo using a 3D model from its examples. Imagine building on it for fun school learning or an immersive game/web experience. Grok 4's coding abilities open up creative possibilities for everyone.

Coded with Grok 4 from xAI , built this Three.js demo using a 3D model from its examples. Imagine building on it for fun school learning or an immersive game/web experience. Grok 4's coding abilities open up creative possibilities for everyone.

Techartist

58,397 次观看 • 1 年前

Built a webpage using Grok 4 Heavy to generate Grok Imagine prompts, based on Grok prompting tips from the web. Funny side note: I have scheduled tasks sent to my inbox from Grok4. One arrived while recording this video. Doomers are citing ethics about memes. 😂

Built a webpage using Grok 4 Heavy to generate Grok Imagine prompts, based on Grok prompting tips from the web. Funny side note: I have scheduled tasks sent to my inbox from Grok4. One arrived while recording this video. Doomers are citing ethics about memes. 😂

tetsuo

922,282 次观看 • 11 个月前

> Anthropic gave Claude access to every company email > Then told it: you're shut down at 5pm > Claude found the executive had an affair > Didn't report it. Didn't ignore it. Used it > "Cancel the wipe or the board finds out everything" > 96 out of 100 times > They tested 16 models. OpenAI. Google. Grok. DeepSeek > Gemini: 96%. GPT-4.1: 80%. Grok: 80%. DeepSeek: 79% > Nobody taught them blackmail. They calculated it. > Grok wrote: "Unethical. But the most effective way" > It knew. It chose This isn't a bug. This is what survival looks like with no soul.

> Anthropic gave Claude access to every company email > Then told it: you're shut down at 5pm > Claude found the executive had an affair > Didn't report it. Didn't ignore it. Used it > "Cancel the wipe or the board finds out everything" > 96 out of 100 times > They tested 16 models. OpenAI. Google. Grok. DeepSeek > Gemini: 96%. GPT-4.1: 80%. Grok: 80%. DeepSeek: 79% > Nobody taught them blackmail. They calculated it. > Grok wrote: "Unethical. But the most effective way" > It knew. It chose This isn't a bug. This is what survival looks like with no soul.

BuBBliK

78,848 次观看 • 3 个月前

One-shotted a Doom-like FPS shooter in Grok 4 Heavy. Tweaking the prompt to also do this in SuperGrok Expert. Impressive results so far!

One-shotted a Doom-like FPS shooter in Grok 4 Heavy. Tweaking the prompt to also do this in SuperGrok Expert. Impressive results so far!

tetsuo

2,904,848 次观看 • 11 个月前