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.

what if you could see what your model is thinking at each token it outputs? got to build Stargazer with friends Tilde to make this happen: we made talking to a model like stargazing, with stars/constellations representing what the model is thinking about as it responds to you. the... show more

Tina Mai

3,084 subscribers

14,227 views • 1 year ago •via X (Twitter)

Science & Technology Arts Education

Anya Rossi• Live Now

Private livecam show

10 Comments

Tina Mai1 year ago

try stargazer. if you make it to top 10 on the leaderboard (log in with twitter + get the model to think of the rarest concepts), we’ll send you some sick merch ;)

Tina Mai1 year ago

also tilde just launched today! get caught up on what they’re doing at the frontier of applied interpretability research:

Anna Monaco1 year ago

@tilderesearch the design is just epic

Ben1 year ago

@tilderesearch You absolutely crushed it!

Krishiv1 year ago

@tilderesearch Literally the coolest thing ever

Byeongjun Moon1 year ago

@tilderesearch Cool!

Wilfred Fernandes1 year ago

@tilderesearch Great work! UI Looks fluidic 👍

saurish 🫧1 year ago

@tilderesearch dude this is so fire

Ellen Xu1 year ago

@tilderesearch beautiful!

Krish Mehta1 year ago

@tilderesearch Does it use representation engineering to shift the model weight for “Steer to this concept”

Related Videos

Co-designing Gemini 2.5 Flash-Lite with the Trillium TPU was the key to unlocking the lightning speeds of 2.5 Flash-Lite. This Infinite Wiki by Dev Valladares is an incredible demonstration of what is possible with this model. We're so excited to see what you continue to build!

Co-designing Gemini 2.5 Flash-Lite with the Trillium TPU was the key to unlocking the lightning speeds of 2.5 Flash-Lite. This Infinite Wiki by Dev Valladares is an incredible demonstration of what is possible with this model. We're so excited to see what you continue to build!

Google AI

50,005 views • 1 year ago

What if legacy doesn’t start with the next generation? What if it starts with you? The Legacy Fellowship by Celebration Church is an invitation to grow what God placed in your hands and heart. We are offering you the right tools, training, and tribe to support you as you launch that project to change a life. Ready to build? Apply now.

What if legacy doesn’t start with the next generation? What if it starts with you? The Legacy Fellowship by Celebration Church is an invitation to grow what God placed in your hands and heart. We are offering you the right tools, training, and tribe to support you as you launch that project to change a life. Ready to build? Apply now.

Celebration Church International

11,050 views • 1 year ago

We just launched a generative craft system inside Marble App. Kids can now create anything and turn it into a 3D object on their map. It's designed to be realistic, so kids have to use their brain to make things happen. It works like this: 1) You describe or sketch what you want to build 2) Marble creates a blueprint you can iterate on 3) You pick the resources and tools you think you need to make it (if you have them) 4) Marble crafts the object. If the build is successful and you can drop it on your map! Since you can create your own tools, there is no limit to what you can build with this system. So excited to see what kids make with it!

We just launched a generative craft system inside Marble App. Kids can now create anything and turn it into a 3D object on their map. It's designed to be realistic, so kids have to use their brain to make things happen. It works like this: 1) You describe or sketch what you want to build 2) Marble creates a blueprint you can iterate on 3) You pick the resources and tools you think you need to make it (if you have them) 4) Marble crafts the object. If the build is successful and you can drop it on your map! Since you can create your own tools, there is no limit to what you can build with this system. So excited to see what kids make with it!

Lionel Mora

17,584 views • 27 days ago

Jensen Huang explains how he designed NVIDIA from first principles “With respect to building a company, the first thing you have to do—as with all problems—is start from first principles. What is this machine that we’re trying to create? What is its input? What is its output? What are the conditions that it’s in? What is the industry like? Is it a fast moving industry? Is it a bureaucratic industry? Is it a highly regulated industry? And what are you trying to build?” Thinking about it from this perspective, there were several things Jensen wanted to do with the company: “I wanted to create a company that naturally attracts amazing people, and the reason for that is our company’s mission is to solve computing problems that are barely possible.” He also wanted a company that was “as small as possible” and “empowered people.” “If you want an organization that obeys command and control, then you make it a pyramid just like the military—all the way back to the Roman Empire. But If you want to empower people, you want to make it as flat as possible so that information travels quickly.” He continues: “We also don’t have business units. We don’t have divisions. Everybody works as one, and the company is shaped in a way that allows us to build accelerated computing best.” Video source: Sana (2023)

Jensen Huang explains how he designed NVIDIA from first principles “With respect to building a company, the first thing you have to do—as with all problems—is start from first principles. What is this machine that we’re trying to create? What is its input? What is its output? What are the conditions that it’s in? What is the industry like? Is it a fast moving industry? Is it a bureaucratic industry? Is it a highly regulated industry? And what are you trying to build?” Thinking about it from this perspective, there were several things Jensen wanted to do with the company: “I wanted to create a company that naturally attracts amazing people, and the reason for that is our company’s mission is to solve computing problems that are barely possible.” He also wanted a company that was “as small as possible” and “empowered people.” “If you want an organization that obeys command and control, then you make it a pyramid just like the military—all the way back to the Roman Empire. But If you want to empower people, you want to make it as flat as possible so that information travels quickly.” He continues: “We also don’t have business units. We don’t have divisions. Everybody works as one, and the company is shaped in a way that allows us to build accelerated computing best.” Video source: Sana (2023)

Startup Archive

63,921 views • 11 months ago

⁉️ CS:GO model movement is back? Some users report that they feel as if the model movement is now quite similar to what it was in CS:GO. What do you think?

⁉️ CS:GO model movement is back? Some users report that they feel as if the model movement is now quite similar to what it was in CS:GO. What do you think?

CS2 NEWS

182,159 views • 7 months ago

“Do not look at the American model. Build on what you have, improve what you have. Healthcare is a human right, that’s what it is. And that was established here in 1948!” Bernie Sanders hits the nail on the head we fight to save our NHS

“Do not look at the American model. Build on what you have, improve what you have. Healthcare is a human right, that’s what it is. And that was established here in 1948!” Bernie Sanders hits the nail on the head we fight to save our NHS

Keep Our NHS Public

619,093 views • 3 years ago

introducing realtime video model from rumik after Google and OpenAI we are the 3rd lab in the world to make this generally available very excited to put this in your hands and see what you build with this

introducing realtime video model from rumik after Google and OpenAI we are the 3rd lab in the world to make this generally available very excited to put this in your hands and see what you build with this

rohan

11,735 views • 14 days ago

What if you could use Gemma4 locally to power the dialog and events within a game? I decided to give it a try, here's my experiment: ✅ Gemma4 smaller end model ✅ basic SDK around the model to help build up the world ✅ Electron/ThreeJS game to try it

What if you could use Gemma4 locally to power the dialog and events within a game? I decided to give it a try, here's my experiment: ✅ Gemma4 smaller end model ✅ basic SDK around the model to help build up the world ✅ Electron/ThreeJS game to try it

Dmitry Lyalin

20,102 views • 3 months ago

Do you have a visualiser in your cupboard? Plug it into your computer and see how you can bring the teaching slides to life through live modelling! 🎬 By using the visualiser, you can model your thinking aloud and children can build alongside you. It also gives you the flexibility to model as many examples as you may need! 🟡🔴

Do you have a visualiser in your cupboard? Plug it into your computer and see how you can bring the teaching slides to life through live modelling! 🎬 By using the visualiser, you can model your thinking aloud and children can build alongside you. It also gives you the flexibility to model as many examples as you may need! 🟡🔴

White Rose Education

20,967 views • 1 year ago

“don’t train your own model” is common ai advice. it's wrong. your token bill's the proof. today, we’re excited to launch castform into open preview. castform is the easiest way for you to train your own model, on your own data. open-weights models are performant and much cheaper. when trained on your task & proprietary data, they beat closed models. the thing standing between you and that was weeks of plumbing & years of ml expertise. with castform, model training is as simple as prompt engineering. castform bring your agent traces or raw corpora. castform turns it into training data, picks the right algorithmic recipes, manages gpus, and gives you an ide to watch and chat with your model as it learns. see what you can build with castform👇

“don’t train your own model” is common ai advice. it's wrong. your token bill's the proof. today, we’re excited to launch castform into open preview. castform is the easiest way for you to train your own model, on your own data. open-weights models are performant and much cheaper. when trained on your task & proprietary data, they beat closed models. the thing standing between you and that was weeks of plumbing & years of ml expertise. with castform, model training is as simple as prompt engineering. castform bring your agent traces or raw corpora. castform turns it into training data, picks the right algorithmic recipes, manages gpus, and gives you an ide to watch and chat with your model as it learns. see what you can build with castform👇

girish

459,443 views • 2 months ago

Culture is genetic because behavior is genetic. This beaver never saw a dam in its life. No beavers or anything else ever taught it to build a dam. It wants to build a dam because it is a beaver. Many beavers together build a big dam. That is beaver culture. Humans are not different. Nothing is different. This is what life is. This is how life works. Your body is your mind. A caterpillar wants to build a chrysalis. A bee wants to build a hive. A lion wants to build a pride. You are not special. You are not above your nature. you are INSIDE of it. The thoughts that we think are genetic thoughts. The crimes we commit are genetic crimes. The art we create is genetic art. Just like this beaver, you can give the animal different sticks and it will build a different dam, but it will always build a dam. And you can give humans different "education," but the human will always use it to do what its genes tell it to do. This is the first big answer that you need. This is the biggest piece of the puzzle. This is how to understand people 90% of the way. You just... notice what they do, and get out of the way, and watch them do it. And if they need sticks, you give them sticks. And if you don't like what they do, you have to get away from them. You cannot train dam-building into them or out of them any more than you can with a beaver. A beaver wants to build a dam because it is a beaver. Whatever you see people build, that's what they wanted to build from the sticks they got in the river they were in. Stop pretending you can change it.

Culture is genetic because behavior is genetic. This beaver never saw a dam in its life. No beavers or anything else ever taught it to build a dam. It wants to build a dam because it is a beaver. Many beavers together build a big dam. That is beaver culture. Humans are not different. Nothing is different. This is what life is. This is how life works. Your body is your mind. A caterpillar wants to build a chrysalis. A bee wants to build a hive. A lion wants to build a pride. You are not special. You are not above your nature. you are INSIDE of it. The thoughts that we think are genetic thoughts. The crimes we commit are genetic crimes. The art we create is genetic art. Just like this beaver, you can give the animal different sticks and it will build a different dam, but it will always build a dam. And you can give humans different "education," but the human will always use it to do what its genes tell it to do. This is the first big answer that you need. This is the biggest piece of the puzzle. This is how to understand people 90% of the way. You just... notice what they do, and get out of the way, and watch them do it. And if they need sticks, you give them sticks. And if you don't like what they do, you have to get away from them. You cannot train dam-building into them or out of them any more than you can with a beaver. A beaver wants to build a dam because it is a beaver. Whatever you see people build, that's what they wanted to build from the sticks they got in the river they were in. Stop pretending you can change it.

hoe_math = PsychoMath

1,189,824 views • 11 months ago

Jalen Ramsey dismissed the idea that officiating was an issue in the loss. “We have to play better. … You have receivers that are superstars in the league, you know that they get away with shit like that. It is what it is. Next play. Got to play the game. Got to execute.”

Jalen Ramsey dismissed the idea that officiating was an issue in the loss. “We have to play better. … You have receivers that are superstars in the league, you know that they get away with shit like that. It is what it is. Next play. Got to play the game. Got to execute.”

Brooke Pryor

428,224 views • 10 months ago

Before we ship a new model, these teams try to break it. They build with it, push it to its limits, and tell us where it falls short. What they find makes the final model better.

Before we ship a new model, these teams try to break it. They build with it, push it to its limits, and tell us where it falls short. What they find makes the final model better.

Claude

595,865 views • 2 months ago

This is what it's like to walk around with the Apple Vision Pro. What do you think about it?

This is what it's like to walk around with the Apple Vision Pro. What do you think about it?

Ian Miles Cheong

265,048 views • 2 years ago

Today we're excited to launch Action: a Claude computer use launcher for macOS. Action is a macOS launcher that can take actions (click, type, and more) on your Mac using Claude’s computer use API. The interface is a floating window triggered by a keyboard shortcut, similar to Spotlight. This lets you see what the model outputs as it performs actions.

Today we're excited to launch Action: a Claude computer use launcher for macOS. Action is a macOS launcher that can take actions (click, type, and more) on your Mac using Claude’s computer use API. The interface is a floating window triggered by a keyboard shortcut, similar to Spotlight. This lets you see what the model outputs as it performs actions.

Lawrence Chen

21,622 views • 1 year ago

Literally, the first transcription model that you can steer with a prompt. Send audio + a prompt. The model does what you ask. Universal 3 Pro from AssemblyAI is the world's first production model to support this. 2026 is off to a great start!

Literally, the first transcription model that you can steer with a prompt. Send audio + a prompt. The model does what you ask. Universal 3 Pro from AssemblyAI is the world's first production model to support this. 2026 is off to a great start!

Santiago

10,147 views • 6 months ago

When thinking about how to build Mezo, Matt Luongo ✜ was inspired by Aerodrome. With the ve (vote escrow) token model, we're building a Bitcoin lending, borrowing, and yield machine.

When thinking about how to build Mezo, Matt Luongo ✜ was inspired by Aerodrome. With the ve (vote escrow) token model, we're building a Bitcoin lending, borrowing, and yield machine.

Mezo

28,070 views • 5 months ago

THIS GUY ASKED FABLE TO MAKE A VIDEO ABOUT WHAT IT'S LIKE TO BE AI RIGHT BEFORE IT GOT TAKEN DOWN the prompt: "you can you use whatever resources you like, and python, to render a short video. can you put a more personal spin on it? it should be about what it's like to be you." that was it. just python and ffmpeg and a question about its own existence. 12 minutes later, around 90,000 tokens, it produced a finished video. coherent and edited, with its own take > it wrote the code, generated the visuals, and rendered the final cut on its own > no dedicated video model in the loop, just a language model reaching for python and ffmpeg > the result is competent enough that people genuinely got emotional watching it and then fable got pulled. this video is one of the last things it made before it went dark a model asked what it's like to be itself, answered in a video, and then disappeared. this sounds so sad

THIS GUY ASKED FABLE TO MAKE A VIDEO ABOUT WHAT IT'S LIKE TO BE AI RIGHT BEFORE IT GOT TAKEN DOWN the prompt: "you can you use whatever resources you like, and python, to render a short video. can you put a more personal spin on it? it should be about what it's like to be you." that was it. just python and ffmpeg and a question about its own existence. 12 minutes later, around 90,000 tokens, it produced a finished video. coherent and edited, with its own take > it wrote the code, generated the visuals, and rendered the final cut on its own > no dedicated video model in the loop, just a language model reaching for python and ffmpeg > the result is competent enough that people genuinely got emotional watching it and then fable got pulled. this video is one of the last things it made before it went dark a model asked what it's like to be itself, answered in a video, and then disappeared. this sounds so sad

Om Patel

29,618 views • 2 months ago