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.

i built the 1st "world model game engine" it makes games from scratch, all running locally on my Mac M1 - 30fps - 40ms latency - ⁠50 million parameters - trained on 15 min of footage made using only "synthetic" datasets this is the first of its kind, &... show more

hugo

1,690 subscribers

230,658 views • 7 months ago •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

0 Comments

No comments available

Comments from the original post will appear here

Related Videos

this is the first battle royale running locally in a world model it can host as many people as you want - 70million parameters - real-time multiplayer - customisable levels it runs in your browser. you can try it now

this is the first battle royale running locally in a world model it can host as many people as you want - 70million parameters - real-time multiplayer - customisable levels it runs in your browser. you can try it now

hugo

39,797 views • 5 months ago

This is a world model running locally on an RTX 5090. It was built from scratch by a team in India, lucidml. Like modern video games, you can choose between performance and graphics by choosing how many times the neural network is run to generate each frame. Absolutely wild.

This is a world model running locally on an RTX 5090. It was built from scratch by a team in India, lucidml. Like modern video games, you can choose between performance and graphics by choosing how many times the neural network is run to generate each frame. Absolutely wild.

Caleb

44,816 views • 6 months ago

This may look like a game of Counterstrike running slowly, but it actually extraordinary. The entire game is created, frame-by-frame, on my *home computer* by an AI diffusion model in response to my actions. There is no game engine, just a "world model" trained on Counterstrike

This may look like a game of Counterstrike running slowly, but it actually extraordinary. The entire game is created, frame-by-frame, on my home computer by an AI diffusion model in response to my actions. There is no game engine, just a "world model" trained on Counterstrike

Ethan Mollick

376,399 views • 1 year ago

i built the 1st game ever on LingBot 2 world model - fits entirely in one node graph - VLM monitors the state in real time - survive longer →new weapons/monsters - powered by Robbyant at 48 fps on reactor available now via API → one call = a playable world first of its kind & you can build yours now

i built the 1st game ever on LingBot 2 world model - fits entirely in one node graph - VLM monitors the state in real time - survive longer →new weapons/monsters - powered by Robbyant at 48 fps on reactor available now via API → one call = a playable world first of its kind & you can build yours now

hugo

105,404 views • 1 month ago

day 2 of building a self-driving power wheel today i officially trained a self driving model from scratch and deploy it on the car by just simply brute forcing everything, I: > made a remote tele-op and remote data collection app built on LiveKit infra > feat: 60ms e2e latency between the car and inference compute (car and compute in vietnam with singapore sfu) > feat: data is collected on operator side, baking latency into observation space itself (I expect this made the model more robust against latency) > recorded 30 min of data at 30fps and converted the dataset to lerobot (you can check a sample here) > trained a simple ACT model (3 epoch, batch size 8) to drive the car around my house > deployed the model on the car with remote inference the video explains everything shortly reflection: > the model is ofc bad, idt behavior cloning would work at all for such complex task on such small sample size > it did work in some cases where the observation is well within distribution, even generalizes to back the car when it gets stuck up next: > will hack alpamayo (NVIDIA) or comma ’s e2e to somehow fit this > or train with a llm backbone or a locomotion prior to see if it generalizes

day 2 of building a self-driving power wheel today i officially trained a self driving model from scratch and deploy it on the car by just simply brute forcing everything, I: > made a remote tele-op and remote data collection app built on LiveKit infra > feat: 60ms e2e latency between the car and inference compute (car and compute in vietnam with singapore sfu) > feat: data is collected on operator side, baking latency into observation space itself (I expect this made the model more robust against latency) > recorded 30 min of data at 30fps and converted the dataset to lerobot (you can check a sample here) > trained a simple ACT model (3 epoch, batch size 8) to drive the car around my house > deployed the model on the car with remote inference the video explains everything shortly reflection: > the model is ofc bad, idt behavior cloning would work at all for such complex task on such small sample size > it did work in some cases where the observation is well within distribution, even generalizes to back the car when it gets stuck up next: > will hack alpamayo (NVIDIA) or comma ’s e2e to somehow fit this > or train with a llm backbone or a locomotion prior to see if it generalizes

Binh Pham

19,404 views • 5 months ago

Akademiks speaks on the “Grand Wizard” running the Akademiks TV page as an Asian from Australia: Ak says he ain’t ducking no smoke “stop using my own page against me I made it 15 years ago. ‘It’s MY page. I can shut this whole shit down. I gotta bunch of people from all around the world running this page”

Akademiks speaks on the “Grand Wizard” running the Akademiks TV page as an Asian from Australia: Ak says he ain’t ducking no smoke “stop using my own page against me I made it 15 years ago. ‘It’s MY page. I can shut this whole shit down. I gotta bunch of people from all around the world running this page”

CY Chels

364,898 views • 9 months ago

Metroid II best ending running on my original Game Boy that I've had since 1992. I kind of can't believe the amount of games I was able to play through on this. My vision must have been awesome because I can barely see it now.

Metroid II best ending running on my original Game Boy that I've had since 1992. I kind of can't believe the amount of games I was able to play through on this. My vision must have been awesome because I can barely see it now.

Philip Summers

14,646 views • 5 months ago

Scientists have created “the first synthetic species” after 15 years of research costing $40 million dollars It’s a bacteria designed on a computer that “gets its genetic instructions from a synthetic chromosome made by man, not nature, and it's alive. It's alive and self-replicating. It means it can indefinitely grow and make copies of itself.” “Venter believes this is the first baby step in a biological revolution, one in which it will be possible to custom design and reprogram bacteria — DNA is the software of life” This work is considered a foundational milestone in synthetic biology, allegedly proving life could be engineered from scratch

Scientists have created “the first synthetic species” after 15 years of research costing $40 million dollars It’s a bacteria designed on a computer that “gets its genetic instructions from a synthetic chromosome made by man, not nature, and it's alive. It's alive and self-replicating. It means it can indefinitely grow and make copies of itself.” “Venter believes this is the first baby step in a biological revolution, one in which it will be possible to custom design and reprogram bacteria — DNA is the software of life” This work is considered a foundational milestone in synthetic biology, allegedly proving life could be engineered from scratch

Wall Street Apes

206,307 views • 4 months ago

⊹₊⟡⋆ Ｃｏｍｍｉｓｓｉｏｎ⊹₊⟡⋆ I made this animation for the stream waiting scene for my vtube child Feyre 🌿⌛️ 女神VTuber ✧ This kind of animation in comms on my site you like? its the first time i try it ♡

⊹₊⟡⋆ Ｃｏｍｍｉｓｓｉｏｎ⊹₊⟡⋆ I made this animation for the stream waiting scene for my vtube child Feyre 🌿⌛️ 女神VTuber ✧ This kind of animation in comms on my site you like? its the first time i try it ♡

⋆. 𐙚˚࿔ Gwenevyr 𝜗𝜚˚⋆

18,124 views • 1 year ago

we built an action model playing Magic like a human introducing ANA: a normal adversary - 5M parameters - runs on a CPU - 12ms per decision - predicts human pick 90% of the time games today. robots tomorrow. this is step one towards people-shaped policies & you can try it now 👇

we built an action model playing Magic like a human introducing ANA: a normal adversary - 5M parameters - runs on a CPU - 12ms per decision - predicts human pick 90% of the time games today. robots tomorrow. this is step one towards people-shaped policies & you can try it now 👇

hugo

18,723 views • 1 month ago

Apple built a large foundation model and fine-tuned it on multiple tasks. But they are doing something very clever: They load a single model in memory and use different adapters to specialize the model on the fly. I recorded a video to show you how to write the code to do the same thing Apple is doing. I explain everything step by step. Here is what I'll show you in the video: 1. We'll load two datasets 2. Then load a large model 3. Then, we'll fine-tune the model on both datasets I'll use LoRA to fine-tune the model. This process creates two small adapters, each specializing in solving one of the datasets. The base model's original parameters will remain unchanged. From here: 4. We'll generate a list of tasks 5. We'll load the correct adapter to solve each task The large model I'm using needs 346 MB of memory, but I only need to load it once. Each adapter is only 2.7 MB. I only need to load the base model once and pair it with any of the fine-tuned adapters. Minimum memory footprint and I can solve multiple tasks. Hope this helps!

Apple built a large foundation model and fine-tuned it on multiple tasks. But they are doing something very clever: They load a single model in memory and use different adapters to specialize the model on the fly. I recorded a video to show you how to write the code to do the same thing Apple is doing. I explain everything step by step. Here is what I'll show you in the video: 1. We'll load two datasets 2. Then load a large model 3. Then, we'll fine-tune the model on both datasets I'll use LoRA to fine-tune the model. This process creates two small adapters, each specializing in solving one of the datasets. The base model's original parameters will remain unchanged. From here: 4. We'll generate a list of tasks 5. We'll load the correct adapter to solve each task The large model I'm using needs 346 MB of memory, but I only need to load it once. Each adapter is only 2.7 MB. I only need to load the base model once and pair it with any of the fine-tuned adapters. Minimum memory footprint and I can solve multiple tasks. Hope this helps!

Santiago

84,747 views • 2 years ago

I finally finished my model in genshin! I made the it myself!! ☀️ Let me know if you’re interested in a model of your OC like this (hoyo games only)! You need the game on pc for custom models to work!

I finally finished my model in genshin! I made the it myself!! ☀️ Let me know if you’re interested in a model of your OC like this (hoyo games only)! You need the game on pc for custom models to work!

Soleil ☀️

180,113 views • 11 months ago

🎙️ INTERVIEW 🎙️ Avowed developers confirm 30FPS was the target FPS from the start of development “Since the game is a first person single player game you don't need a 60fps mode” It could be 30FPS locked on Xbox How do you all feel about this?💚🎮

🎙️ INTERVIEW 🎙️ Avowed developers confirm 30FPS was the target FPS from the start of development “Since the game is a first person single player game you don't need a 60fps mode” It could be 30FPS locked on Xbox How do you all feel about this?💚🎮

VGT Gaming News

166,926 views • 2 years ago

i barely write ad scripts from scratch anymore. i plugged WinningHunter’s MCP into my workflow so i can ask it what’s working on the algorithm and turn that data into new ad concepts on the fly. this entire ad was researched + scripted using their MCP. the brief was literally: “what’s working right now for women 55+?” it pulled the patterns, angles + ads getting traction and we built the concept around that data. made this out of it for a client. this is where ad research is going. give it a try here:

i barely write ad scripts from scratch anymore. i plugged WinningHunter’s MCP into my workflow so i can ask it what’s working on the algorithm and turn that data into new ad concepts on the fly. this entire ad was researched + scripted using their MCP. the brief was literally: “what’s working right now for women 55+?” it pulled the patterns, angles + ads getting traction and we built the concept around that data. made this out of it for a client. this is where ad research is going. give it a try here:

Franky Shaw

17,051 views • 2 days ago

AI can now build an entire world Using Grok 4.5 + Grok Build in Unreal Engine 5.8, an autonomous agent created a detailed cyberpunk street from scratch. → 36.5 minutes → 10.75 million tokens → Only $12.40. It checked its own work, fixed mistakes, and saved 30+ checkpoints. The future of 3D creation is here. Writer: Val

AI can now build an entire world Using Grok 4.5 + Grok Build in Unreal Engine 5.8, an autonomous agent created a detailed cyberpunk street from scratch. → 36.5 minutes → 10.75 million tokens → Only $12.40. It checked its own work, fixed mistakes, and saved 30+ checkpoints. The future of 3D creation is here. Writer: Val

Mario Nawfal

49,074 views • 1 month ago

I applied Matt's prompt, ran it for a little over ten hours. About 1.3 million tokens on Opus 5. It created the game. I also made a video of it using Opus 5. Here's the video (and in my reply to this is a link to the actual game so you can play).

I applied Matt's prompt, ran it for a little over ten hours. About 1.3 million tokens on Opus 5. It created the game. I also made a video of it using Opus 5. Here's the video (and in my reply to this is a link to the actual game so you can play).

James Altucher

128,128 views • 1 month ago

OpenAI has finally released GPT OSS 20B!! This is probably the most powerful model you can run locally on a laptop. → Only 3.6B active parameters → Adjustable "Reasoning Effort" → Works with 16GB of RAM! Quick steps to use it 100% free and offline:

OpenAI has finally released GPT OSS 20B!! This is probably the most powerful model you can run locally on a laptop. → Only 3.6B active parameters → Adjustable "Reasoning Effort" → Works with 16GB of RAM! Quick steps to use it 100% free and offline:

Paul Couvert

595,738 views • 1 year ago

I can’t believe this is running on my iPhone My first Unreal Engine game on iOS… and it’s Hermes Agora A 3D command center where I can see my Hermes agents, walk around the office, and interact with them from my phone My agents now have a world I can carry in my pocket

I can’t believe this is running on my iPhone My first Unreal Engine game on iOS… and it’s Hermes Agora A 3D command center where I can see my Hermes agents, walk around the office, and interact with them from my phone My agents now have a world I can carry in my pocket

Luke The Dev

37,617 views • 2 months ago

This is my first ever game that I made in 2 weeks, it's about my internalized fear of women and relationships. I'm expecting to make a million dollars from sales. The game is "I Just Want to be Single!!" and you can play the demo right now on Steam

Sensitive content

This is my first ever game that I made in 2 weeks, it's about my internalized fear of women and relationships. I'm expecting to make a million dollars from sales. The game is "I Just Want to be Single!!" and you can play the demo right now on Steam

Tsundere Studio

55,274 views • 2 years ago