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。

I love this opensource project! Someone just put a 28.9M parameter AI model on an ESP32-S3. A chip that costs only a few dollars. No cloud. No API. No internet. It runs locally at ~9.5 tokens/s and can generate stories on a tiny screen. This changes the way we... show more

Steven Cheng

4,552 subscribers

100,486 次观看 • 2 天前 •via X (Twitter)

科学技术新闻政治

Anya Rossi• Live Now

Private livecam show

0 条评论

暂无评论

原始帖子的评论将显示在这里

相关视频

forget the $699 AI pins. this $8 chip just shattered the barrier for local AI hardware. a developer just forced a 28.9 million-parameter LLM onto a standard ESP32-S3 microcontroller. it costs roughly 8 dollars, runs completely offline, and draws the power of a single LED. conventional wisdom said a model of this size simply would not fit. the chip only has 512 KB of fast SRAM and 16 MB of flash. the breakthrough is architectural. the developer moved the bulk of the embedding table into flash memory and memory-mapped it. the chip only needs to pull about 450 bytes per token, keeping the active working memory inside the fast SRAM. this means you can now embed a capable language model into a physical node for the price of two coffees. and we are already seeing the beginnings of this custom physical hardware. in the video, a creator built a minimalist voice-controlled universal remote using an ESP32. it captures voice and remotely controls the computer over bluetooth LE. he simply says "open chrome and open 20 new tabs", and the custom hardware executes it instantly. we have spent years watching model sizes explode upward. but the true frontier is the opposite direction. when an eight-dollar chip can power offline intelligence and custom physical interfaces, AI becomes local infrastructure rather than a cloud service.

forget the $699 AI pins. this $8 chip just shattered the barrier for local AI hardware. a developer just forced a 28.9 million-parameter LLM onto a standard ESP32-S3 microcontroller. it costs roughly 8 dollars, runs completely offline, and draws the power of a single LED. conventional wisdom said a model of this size simply would not fit. the chip only has 512 KB of fast SRAM and 16 MB of flash. the breakthrough is architectural. the developer moved the bulk of the embedding table into flash memory and memory-mapped it. the chip only needs to pull about 450 bytes per token, keeping the active working memory inside the fast SRAM. this means you can now embed a capable language model into a physical node for the price of two coffees. and we are already seeing the beginnings of this custom physical hardware. in the video, a creator built a minimalist voice-controlled universal remote using an ESP32. it captures voice and remotely controls the computer over bluetooth LE. he simply says "open chrome and open 20 new tabs", and the custom hardware executes it instantly. we have spent years watching model sizes explode upward. but the true frontier is the opposite direction. when an eight-dollar chip can power offline intelligence and custom physical interfaces, AI becomes local infrastructure rather than a cloud service.

ard

417,806 次观看 • 10 天前

Holy shit... Microsoft open sourced an inference framework that runs a 100B parameter LLM on a single CPU. It's called BitNet. And it does what was supposed to be impossible. No GPU. No cloud. No $10K hardware setup. Just your laptop running a 100-billion parameter model at human reading speed. Here's how it works: Every other LLM stores weights in 32-bit or 16-bit floats. BitNet uses 1.58 bits. Weights are ternary just -1, 0, or +1. That's it. No floats. No expensive matrix math. Pure integer operations your CPU was already built for. The result: - 100B model runs on a single CPU at 5-7 tokens/second - 2.37x to 6.17x faster than llama.cpp on x86 - 82% lower energy consumption on x86 CPUs - 1.37x to 5.07x speedup on ARM (your MacBook) - Memory drops by 16-32x vs full-precision models The wildest part: Accuracy barely moves. BitNet b1.58 2B4T their flagship model was trained on 4 trillion tokens and benchmarks competitively against full-precision models of the same size. The quantization isn't destroying quality. It's just removing the bloat. What this actually means: - Run AI completely offline. Your data never leaves your machine - Deploy LLMs on phones, IoT devices, edge hardware - No more cloud API bills for inference - AI in regions with no reliable internet The model supports ARM and x86. Works on your MacBook, your Linux box, your Windows machine. 27.4K GitHub stars. 2.2K forks. Built by Microsoft Research. 100% Open Source. MIT License.

Holy shit... Microsoft open sourced an inference framework that runs a 100B parameter LLM on a single CPU. It's called BitNet. And it does what was supposed to be impossible. No GPU. No cloud. No $10K hardware setup. Just your laptop running a 100-billion parameter model at human reading speed. Here's how it works: Every other LLM stores weights in 32-bit or 16-bit floats. BitNet uses 1.58 bits. Weights are ternary just -1, 0, or +1. That's it. No floats. No expensive matrix math. Pure integer operations your CPU was already built for. The result: - 100B model runs on a single CPU at 5-7 tokens/second - 2.37x to 6.17x faster than llama.cpp on x86 - 82% lower energy consumption on x86 CPUs - 1.37x to 5.07x speedup on ARM (your MacBook) - Memory drops by 16-32x vs full-precision models The wildest part: Accuracy barely moves. BitNet b1.58 2B4T their flagship model was trained on 4 trillion tokens and benchmarks competitively against full-precision models of the same size. The quantization isn't destroying quality. It's just removing the bloat. What this actually means: - Run AI completely offline. Your data never leaves your machine - Deploy LLMs on phones, IoT devices, edge hardware - No more cloud API bills for inference - AI in regions with no reliable internet The model supports ARM and x86. Works on your MacBook, your Linux box, your Windows machine. 27.4K GitHub stars. 2.2K forks. Built by Microsoft Research. 100% Open Source. MIT License.

Guri Singh

2,180,357 次观看 • 4 个月前

🤯 What, Chrome has an AI summarizer API built in now?? - No API keys, no internet. Runs locally - Downloads the model on-demand - Available as a global in the latest version of Chrome. Works in chrome extensions too I have so many chrome extension ideas now...

🤯 What, Chrome has an AI summarizer API built in now?? - No API keys, no internet. Runs locally - Downloads the model on-demand - Available as a global in the latest version of Chrome. Works in chrome extensions too I have so many chrome extension ideas now...

Ben Holmes

89,797 次观看 • 11 个月前

Cancelled ChatGPT -> Built JARVIS -> Pays $0 -> it works offline + it's smarter than the $20/month version. No WiFi needed, no cloud, no API keys, no rate limits, no queues, no $20/month just to ask a server in Virginia for the weather. Just a local model running directly on the laptop hardware, voice activated, system integrated, controlling apps, answering questions, doing the work. Iron Man had JARVIS embedded in his suit, this guy has it embedded in his MacBook and it works on a plane, in a basement, on a remote cabin with zero signal. OpenAI is burning $700,000 a day on infrastructure to deliver something this guy runs for free. Anthropic charges $200/month for unlimited Claude access, microsoft built Copilot into every product they sell. This guy skipped all of it, downloaded a model and made his laptop the smartest device in the room. No subscription. No login. No internet. No data sent anywhere ever. The most powerful AI assistant on earth is now the one running locally on hardware you already own. ChatGPT charges you to think slower, he pays nothing and thinks alone, he made it himself.

Cancelled ChatGPT -> Built JARVIS -> Pays $0 -> it works offline + it's smarter than the $20/month version. No WiFi needed, no cloud, no API keys, no rate limits, no queues, no $20/month just to ask a server in Virginia for the weather. Just a local model running directly on the laptop hardware, voice activated, system integrated, controlling apps, answering questions, doing the work. Iron Man had JARVIS embedded in his suit, this guy has it embedded in his MacBook and it works on a plane, in a basement, on a remote cabin with zero signal. OpenAI is burning $700,000 a day on infrastructure to deliver something this guy runs for free. Anthropic charges $200/month for unlimited Claude access, microsoft built Copilot into every product they sell. This guy skipped all of it, downloaded a model and made his laptop the smartest device in the room. No subscription. No login. No internet. No data sent anywhere ever. The most powerful AI assistant on earth is now the one running locally on hardware you already own. ChatGPT charges you to think slower, he pays nothing and thinks alone, he made it himself.

Defileo🔮

154,009 次观看 • 3 个月前

Gemma 4 looks at a parking lot. Decides what to ask. Calls SAM 3.1. "Segment all vehicles." 64 found. "Now just the white ones." 23 found. One model reasoning and orchestrating. One model executing. Both running locally on a MacBook. MLX. No cloud. No API.

Gemma 4 looks at a parking lot. Decides what to ask. Calls SAM 3.1. "Segment all vehicles." 64 found. "Now just the white ones." 23 found. One model reasoning and orchestrating. One model executing. Both running locally on a MacBook. MLX. No cloud. No API.

Maziyar PANAHI

594,086 次观看 • 3 个月前

Gemma 4 watches raw video. Understands the scene. Then prompts SAM 3 to segment and RF-DETR to track. One AI directing two others. Fighter jets. Crowds. Aerial defense footage. All three models running locally on a MacBook. No cloud. What scene should I point this at next?

Gemma 4 watches raw video. Understands the scene. Then prompts SAM 3 to segment and RF-DETR to track. One AI directing two others. Fighter jets. Crowds. Aerial defense footage. All three models running locally on a MacBook. No cloud. What scene should I point this at next?

Maziyar PANAHI

369,767 次观看 • 4 个月前

NVIDIA just made paying for AI feel optional. Open model, a million tokens of context, free tier with no per-token cost, runs on your own hardware. Entire codebases, whole data rooms, a year of chat logs, all swallowed in one prompt. No chunking, no RAG, no rate limit theater. The closed-AI premium has 90 days to defend itself. Bookmark this and come back. Open beat closed. Again.

NVIDIA just made paying for AI feel optional. Open model, a million tokens of context, free tier with no per-token cost, runs on your own hardware. Entire codebases, whole data rooms, a year of chat logs, all swallowed in one prompt. No chunking, no RAG, no rate limit theater. The closed-AI premium has 90 days to defend itself. Bookmark this and come back. Open beat closed. Again.

shmidt

295,692 次观看 • 1 个月前

NVIDIA might have just declared war on the cloud GPU business For years, AI builders had one option Rent compute Pay every month Watch the bill grow every time usage increased Now NVIDIA is putting serious AI hardware directly on people's desks Small enough to fit next to a monitor Powerful enough to run workloads that used to require expensive cloud infrastructure That's why this launch is getting so much attention The real story isn't the hardware specs It's the business model shift Every month, developers send money to cloud providers for inference, testing, fine-tuning and AI applications The question nobody can answer yet is what happens if enough developers decide they'd rather buy infrastructure once than rent it forever Because if local AI hardware keeps getting more powerful, the economics start changing very quickly Cloud providers built empires on renting access to compute NVIDIA is betting more people will eventually want to own it And that's a much bigger story than a new piece of hardware sitting on a desk

NVIDIA might have just declared war on the cloud GPU business For years, AI builders had one option Rent compute Pay every month Watch the bill grow every time usage increased Now NVIDIA is putting serious AI hardware directly on people's desks Small enough to fit next to a monitor Powerful enough to run workloads that used to require expensive cloud infrastructure That's why this launch is getting so much attention The real story isn't the hardware specs It's the business model shift Every month, developers send money to cloud providers for inference, testing, fine-tuning and AI applications The question nobody can answer yet is what happens if enough developers decide they'd rather buy infrastructure once than rent it forever Because if local AI hardware keeps getting more powerful, the economics start changing very quickly Cloud providers built empires on renting access to compute NVIDIA is betting more people will eventually want to own it And that's a much bigger story than a new piece of hardware sitting on a desk

beamnxw ./

30,361 次观看 • 2 个月前

ANTHROPIC AND OPENAI BOTH MISSED THIS. Someone just open-sourced one SDK that gives AI agents a mouse, keyboard, and screen across macOS, Linux, and Windows. It's called CUA. Same code drives a local VM or a cloud sandbox regardless of the OS. Your agent sees the screen, clicks, types, and runs shell commands like a real user. > Benchmark against OSWorld and ScreenSpot > Export trajectories for model training > Near-native performance on Apple Silicon No separate SDK per OS. No infrastructure headache. 100% Free. Open Source.

ANTHROPIC AND OPENAI BOTH MISSED THIS. Someone just open-sourced one SDK that gives AI agents a mouse, keyboard, and screen across macOS, Linux, and Windows. It's called CUA. Same code drives a local VM or a cloud sandbox regardless of the OS. Your agent sees the screen, clicks, types, and runs shell commands like a real user. > Benchmark against OSWorld and ScreenSpot > Export trajectories for model training > Near-native performance on Apple Silicon No separate SDK per OS. No infrastructure headache. 100% Free. Open Source.

Simplifying AI

31,222 次观看 • 17 天前

JENSEN HUANG UNVEILED A BOARD THAT RUNS 1 TRILLION PARAMETER AI MODELS. THE $249 NVIDIA BOX UNDER YOUR DESK KILLS A $200/MONTH AI BILL FOR $5 IN ELECTRICITY jensen held it up on stage with one hand and called it the architecture that runs the future of ai. that same technology now ships in a $249 box smaller than your wallet the jetson orin nano super pulls 7-25 watts and does 67 trillion ai operations per second. llama 3, mistral and deepseek run locally with no api fees and no data leaving your machine most developers pay $2,400 a year across chatgpt, openai api, claude pro and cursor. the jetson costs $314 in year one and $60 a year after. 2 year savings hit $4,431 install ollama with one command, change one line of code to point at localhost, and every tool built for openai works identically. zero rewrites, zero rate limits cloud subscriptions keep getting more expensive and rate limits keep getting tighter. the people who own the box in 2026 are going to look very far ahead in 2028 bookmark this and read the article below

JENSEN HUANG UNVEILED A BOARD THAT RUNS 1 TRILLION PARAMETER AI MODELS. THE $249 NVIDIA BOX UNDER YOUR DESK KILLS A $200/MONTH AI BILL FOR $5 IN ELECTRICITY jensen held it up on stage with one hand and called it the architecture that runs the future of ai. that same technology now ships in a $249 box smaller than your wallet the jetson orin nano super pulls 7-25 watts and does 67 trillion ai operations per second. llama 3, mistral and deepseek run locally with no api fees and no data leaving your machine most developers pay $2,400 a year across chatgpt, openai api, claude pro and cursor. the jetson costs $314 in year one and $60 a year after. 2 year savings hit $4,431 install ollama with one command, change one line of code to point at localhost, and every tool built for openai works identically. zero rewrites, zero rate limits cloud subscriptions keep getting more expensive and rate limits keep getting tighter. the people who own the box in 2026 are going to look very far ahead in 2028 bookmark this and read the article below

starmex

54,448 次观看 • 2 个月前

⚠️ SECURITY: VITALIK BUTERIN DITCHES CLOUD AI, REVEALS FULLY LOCAL LLM SETUP Ethereum co-founder vitalik.eth has moved entirely off cloud AI services and published his full private AI stack in a new blog post. He now runs the Qwen3.5:35B model locally on an Nvidia 5090 laptop, hitting 90 tokens per second. He cited research showing roughly 15% of AI agent skills contain malicious instructions, and that parsing a single malicious webpage could fully compromise an AI assistant. Buterin also open-sourced a messaging daemon requiring human approval for all outbound messages to third parties.

⚠️ SECURITY: VITALIK BUTERIN DITCHES CLOUD AI, REVEALS FULLY LOCAL LLM SETUP Ethereum co-founder vitalik.eth has moved entirely off cloud AI services and published his full private AI stack in a new blog post. He now runs the Qwen3.5:35B model locally on an Nvidia 5090 laptop, hitting 90 tokens per second. He cited research showing roughly 15% of AI agent skills contain malicious instructions, and that parsing a single malicious webpage could fully compromise an AI assistant. Buterin also open-sourced a messaging daemon requiring human approval for all outbound messages to third parties.

BSCN

25,445 次观看 • 4 个月前

THIS $2 AI VISUAL SETUP JUST MADE A $250/MONTH CREATIVE STACK LOOK STUPID he moves his hand, the particles react, and the whole scene updates in real time. no cloud render farm, no paid visual API, no expensive plugin chain quietly eating money every month most people still think ai visuals require a stack of $39 tools, $89 subscriptions, and constant API usage. but here the loop is simple: TouchDesigner handles the visuals, a local model handles the logic, and the laptop does the rest the important part is not that the particles look cool. the important part is that the “brain” behind the visual no longer has to live on someone else’s server once that moves local, the monthly bill falls off a cliff this is probably how a lot of small studios start building visuals soon: not by renting 6 tools forever, but by owning one system that runs almost for free

THIS $2 AI VISUAL SETUP JUST MADE A $250/MONTH CREATIVE STACK LOOK STUPID he moves his hand, the particles react, and the whole scene updates in real time. no cloud render farm, no paid visual API, no expensive plugin chain quietly eating money every month most people still think ai visuals require a stack of $39 tools, $89 subscriptions, and constant API usage. but here the loop is simple: TouchDesigner handles the visuals, a local model handles the logic, and the laptop does the rest the important part is not that the particles look cool. the important part is that the “brain” behind the visual no longer has to live on someone else’s server once that moves local, the monthly bill falls off a cliff this is probably how a lot of small studios start building visuals soon: not by renting 6 tools forever, but by owning one system that runs almost for free

Ridark

15,948 次观看 • 2 个月前

Elon Musk just identified the next crisis in AI. It’s not a shortage. It’s an unusable surplus. Musk: “By the end of this year, chip production will outpace the ability to turn chips on.” For three years the world was starved for silicon. Every lab, every government, every company racing to secure the chips that determine who wins the AI era. That bottleneck is ending. A new one is replacing it. Musk: “The chips are going to be piling up and not be able to be turned on.” Billions of dollars of the most advanced AI hardware ever built. Sitting dark. Not because the chips don’t work. Because there isn’t enough electricity to run them. You can’t print a power plant the way you print a chip. The fabrication plants scaled. The grid didn’t. And now the most valuable hardware in history is about to hit a wall that no amount of capital can instantly solve. Compute is about to become abundant. Electricity is about to become the most valuable commodity on earth. Three years obsessing over silicon yields. Physics doesn’t care about your chip architecture if your data center can’t pull enough megawatts. The war isn’t about who can manufacture the most silicon anymore. It’s about who has the raw power to plug it in. Whoever solves energy first doesn’t just win. They own the infrastructure everyone else needs to compete. The losers stack useless chips in warehouses waiting for power that never arrives. We built a trillion dollar engine and forgot the fuel. That’s the AI race right now.

Elon Musk just identified the next crisis in AI. It’s not a shortage. It’s an unusable surplus. Musk: “By the end of this year, chip production will outpace the ability to turn chips on.” For three years the world was starved for silicon. Every lab, every government, every company racing to secure the chips that determine who wins the AI era. That bottleneck is ending. A new one is replacing it. Musk: “The chips are going to be piling up and not be able to be turned on.” Billions of dollars of the most advanced AI hardware ever built. Sitting dark. Not because the chips don’t work. Because there isn’t enough electricity to run them. You can’t print a power plant the way you print a chip. The fabrication plants scaled. The grid didn’t. And now the most valuable hardware in history is about to hit a wall that no amount of capital can instantly solve. Compute is about to become abundant. Electricity is about to become the most valuable commodity on earth. Three years obsessing over silicon yields. Physics doesn’t care about your chip architecture if your data center can’t pull enough megawatts. The war isn’t about who can manufacture the most silicon anymore. It’s about who has the raw power to plug it in. Whoever solves energy first doesn’t just win. They own the infrastructure everyone else needs to compete. The losers stack useless chips in warehouses waiting for power that never arrives. We built a trillion dollar engine and forgot the fuel. That’s the AI race right now.

Dustin

705,429 次观看 • 5 个月前

Elon announced Macrohard, a joint xAI-Tesla project where Grok acts as the brain directing a real-time AI agent that can process your screen and actions live. Runs on a $650 Tesla chip. He said it can emulate entire companies. The name alone is an all-time troll of Microsoft.

Elon announced Macrohard, a joint xAI-Tesla project where Grok acts as the brain directing a real-time AI agent that can process your screen and actions live. Runs on a $650 Tesla chip. He said it can emulate entire companies. The name alone is an all-time troll of Microsoft.

Mario Nawfal

309,805 次观看 • 4 个月前

This project is made by humans, ive seen the comment quite a bit but this is not AI at all we drew everything from scratch and i painted the Glass mosaics, I'm working on getting a composer and Sound team on board next. Thank you for supporting our Indie animation project!

This project is made by humans, ive seen the comment quite a bit but this is not AI at all we drew everything from scratch and i painted the Glass mosaics, I'm working on getting a composer and Sound team on board next. Thank you for supporting our Indie animation project!

poppihin I THE EBONY WITCH

48,493 次观看 • 3 个月前

Most video tools can generate clips. Very few can maintain identity. That has been the real bottleneck in AI video creation. Kling O1 changes that. For the first time, creators can carry a character, style, and visual language across scenes without constant fixes. You can reference past clips, assets, or images and the output stays consistently on-model. No visual drift. No rework loops. No “this doesn’t look like the last shot” moments. It feels less like prompting a tool and more like working with a creative collaborator that remembers context. The impact is practical, not theoretical: → Faster production cycles → Lower iteration costs → Noticeably higher output quality This is what mature AI tooling looks like. Not louder features. Not bigger claims. Just reliability where it actually matters. Consistency is no longer the problem.

Most video tools can generate clips. Very few can maintain identity. That has been the real bottleneck in AI video creation. Kling O1 changes that. For the first time, creators can carry a character, style, and visual language across scenes without constant fixes. You can reference past clips, assets, or images and the output stays consistently on-model. No visual drift. No rework loops. No “this doesn’t look like the last shot” moments. It feels less like prompting a tool and more like working with a creative collaborator that remembers context. The impact is practical, not theoretical: → Faster production cycles → Lower iteration costs → Noticeably higher output quality This is what mature AI tooling looks like. Not louder features. Not bigger claims. Just reliability where it actually matters. Consistency is no longer the problem.

Darshal Jaitwar

141,038 次观看 • 6 个月前

Holy sh!t ! OpenAI will have their custom inference chips ready in just a few months and deployed at scale by the end of the year! 🤯 Training chip = The heavy lifters that require massive amounts of data and power to build and teach the AI models from scratch. Inference chip = The specialized, highly efficient chips that actually run the AI and generate the answers in real-time when you use it. This is going to help OpenAI drastically cut down their massive compute costs, speed up model reasoning times, and finally break free from relying entirely on Nvidia to scale their operations.

Holy sh!t ! OpenAI will have their custom inference chips ready in just a few months and deployed at scale by the end of the year! 🤯 Training chip = The heavy lifters that require massive amounts of data and power to build and teach the AI models from scratch. Inference chip = The specialized, highly efficient chips that actually run the AI and generate the answers in real-time when you use it. This is going to help OpenAI drastically cut down their massive compute costs, speed up model reasoning times, and finally break free from relying entirely on Nvidia to scale their operations.

Chris

60,278 次观看 • 4 个月前

Got GLM-5.2 running on my Mac Studio via llama.cpp, the reasoning behind all my medical agentic workflows. It orchestrates a swarm of tiny on-device OpenMed experts: oncology, meds, labs. No cloud, no rate limits, nobody can take it away. AI must be owned, not rented.

Got GLM-5.2 running on my Mac Studio via llama.cpp, the reasoning behind all my medical agentic workflows. It orchestrates a swarm of tiny on-device OpenMed experts: oncology, meds, labs. No cloud, no rate limits, nobody can take it away. AI must be owned, not rented.

Maziyar PANAHI

87,461 次观看 • 1 个月前

1-bit Kimi K3 performs at Opus 5 level on 3D physics! We ran our Atomic Chat quant of Kimi K3 locally on 4x B200 against three cloud models and gave them all the same task, to build a giant anvil drop test as a single HTML file with real physics Outputs: K3 1bit (local): 15.8K tokens, $0 API cost Kimi K3 (API): 15.3K tokens, $0.30 API cost Opus 5: 22.8K tokens, $0.77 API cost GPT 5.6: 14.5K tokens, $0.72 API cost All four got the physics right. But only Kimi made a working winch. The drum turns and the chain drags the flat car off the pad. Opus 5 drew the most detail, road markings and sparks on the hit. And you can run a model at this level on your own box now. That still feels insane to us

1-bit Kimi K3 performs at Opus 5 level on 3D physics! We ran our Atomic Chat quant of Kimi K3 locally on 4x B200 against three cloud models and gave them all the same task, to build a giant anvil drop test as a single HTML file with real physics Outputs: K3 1bit (local): 15.8K tokens, $0 API cost Kimi K3 (API): 15.3K tokens, $0.30 API cost Opus 5: 22.8K tokens, $0.77 API cost GPT 5.6: 14.5K tokens, $0.72 API cost All four got the physics right. But only Kimi made a working winch. The drum turns and the chain drags the flat car off the pad. Opus 5 drew the most detail, road markings and sparks on the hit. And you can run a model at this level on your own box now. That still feels insane to us

atomic.chat

51,542 次观看 • 3 天前