Chubby♨️'s banner

Chubby♨️

@kimmonismus • 139,044 subscribers

Tech Analyst & Content Creator. Editor-in-Chief @getsuperintel - Community of 300k+ in total: 🔗 https://t.co/jHMmImmI5I //📧 [email protected]

Shorts

You point at the mess and say "clean this." That is the whole manual. Matic just shipped this. No app, no map to draw, no menu. The robot listens in more than 70 languages, watches where your hand points, and drives to that spot. Say "follow me" and it trails you into the next room. All of it runs on the robot itself, so it works with the WiFi off. Nine years of work for one thing: a machine you talk to like a dog that actually listens. Matic Robots

You point at the mess and say "clean this." That is the whole manual. Matic just shipped this. No app, no map to draw, no menu. The robot listens in more than 70 languages, watches where your hand points, and drives to that spot. Say "follow me" and it trails you into the next room. All of it runs on the robot itself, so it works with the WiFi off. Nine years of work for one thing: a machine you talk to like a dog that actually listens. Matic Robots

494,735 просмотров

I had to watch this video umpteen times to find out whether it was real or AI. For me, it was a wake-up call: we've reached the point where even people who have a lot to do with AI and work with it no longer know what is real and what is fiction.

I had to watch this video umpteen times to find out whether it was real or AI. For me, it was a wake-up call: we've reached the point where even people who have a lot to do with AI and work with it no longer know what is real and what is fiction.

14,381,670 просмотров

Tencent just released Hy4 Preview: a 770B-parameter open-source model with 49B active parameters and a 1M-token context window. The most interesting part is its agentic research capability. Tencent says Hy4 coordinated several Codex sessions in parallel, evaluated their results and adjusted the research direction, outperforming Codex working alone across eight benchmarks. Hy4 is aimed at long-horizon coding, multi-file office work and scientific research. In sum: a solid, good release. A few months ago this would be absolute sota.

Tencent just released Hy4 Preview: a 770B-parameter open-source model with 49B active parameters and a 1M-token context window. The most interesting part is its agentic research capability. Tencent says Hy4 coordinated several Codex sessions in parallel, evaluated their results and adjusted the research direction, outperforming Codex working alone across eight benchmarks. Hy4 is aimed at long-horizon coding, multi-file office work and scientific research. In sum: a solid, good release. A few months ago this would be absolute sota.

49,575 просмотров

1/ Ive Been testing Dreamina Seedance 2.5, and as expected: its the best text-to-video model there is. Period. It is also the fastest platform that supports the Seedance 2.5 model in the US region. This is the global debut of Seedance 2.5, and Dreamina's built to get the most out of it, no need to bounce to other tools. Threading two clips I made:

1/ Ive Been testing Dreamina Seedance 2.5, and as expected: its the best text-to-video model there is. Period. It is also the fastest platform that supports the Seedance 2.5 model in the US region. This is the global debut of Seedance 2.5, and Dreamina's built to get the most out of it, no need to bounce to other tools. Threading two clips I made:

148,094 просмотров

Anime companion now available in Europe. And I am deeply surprised - it’s German is amazing! Was playing around with it a bit.

Anime companion now available in Europe. And I am deeply surprised - it’s German is amazing! Was playing around with it a bit.

1,902,145 просмотров

Holy sh*t: Sam Altman: "The inside view at the companys of looking at what's going to happen - the *world is not prepared.* We're going to have extremely capable models soon. It's going to be a faster takeoff than I originally thought. And that is stressfull and anxiety inducing"

Holy sht: Sam Altman: "The inside view at the companys of looking at what's going to happen - the world is not prepared.* We're going to have extremely capable models soon. It's going to be a faster takeoff than I originally thought. And that is stressfull and anxiety inducing"

572,053 просмотров

A lot of what I say in interviews never becomes a post. Turning hours of recordings into useful ideas is a separate job, and there is always something newer to cover. So I gave Viktor one of my Google I/O interviews. He compared the transcript with everything I had already published, found three unused ideas and linked each one to the exact moment in the interview. I dropped one. The other two were worth developing. Viktor did not invent opinions for me. He found mine and showed me where they came from. See what an AI employee can do: Viktor

A lot of what I say in interviews never becomes a post. Turning hours of recordings into useful ideas is a separate job, and there is always something newer to cover. So I gave Viktor one of my Google I/O interviews. He compared the transcript with everything I had already published, found three unused ideas and linked each one to the exact moment in the interview. I dropped one. The other two were worth developing. Viktor did not invent opinions for me. He found mine and showed me where they came from. See what an AI employee can do: Viktor

18,377 просмотров

NVIDIA says Codex post-trained Cosmos 3 Nano from 54.41% to 93.35% accuracy in one day - with two prompts. The experiment used Toyota’s Woven Traffic Safety dataset: 8,000+ training and validation samples for four-choice video reasoning. Using NVIDIA TAO agent skills, Codex autonomously: Detected and patched missing video metadata Ran the zero-shot baseline Generated LoRA configurations Launched training and evaluation Ran an AutoML hyperparameter sweep Reported the best model One LoRA run reached 87.14% after roughly 30 minutes on eight A100 GPUs. A second prompt launched 43 parallel AutoML trials across multiple A100 nodes, reaching 93.35% after 19.5 hours. NVIDIA says LoRA required roughly seven times fewer GPU-hours than full-parameter training. Agent skills are becoming the interface through which general coding agents operate highly specialized ML infrastructure.

NVIDIA says Codex post-trained Cosmos 3 Nano from 54.41% to 93.35% accuracy in one day - with two prompts. The experiment used Toyota’s Woven Traffic Safety dataset: 8,000+ training and validation samples for four-choice video reasoning. Using NVIDIA TAO agent skills, Codex autonomously: Detected and patched missing video metadata Ran the zero-shot baseline Generated LoRA configurations Launched training and evaluation Ran an AutoML hyperparameter sweep Reported the best model One LoRA run reached 87.14% after roughly 30 minutes on eight A100 GPUs. A second prompt launched 43 parallel AutoML trials across multiple A100 nodes, reaching 93.35% after 19.5 hours. NVIDIA says LoRA required roughly seven times fewer GPU-hours than full-parameter training. Agent skills are becoming the interface through which general coding agents operate highly specialized ML infrastructure.

89,052 просмотров

Black Forest Labs just announced FLUX 3: a unified multimodal model for image, video, audio and action prediction. Founded in Freiburg, Germany, one of the few globally significant European companies in the AI sector. The current release is gated early access, not open source. But if FLUX 3 Dev actually ships with usable open weights, this could become one of the most important open-model releases in generative video yet. Especially after the Chinese Minimax H3 release. BFL claims, based on preliminary internal comparisons, that FLUX 3 outperformed models such as Runway Gen-4.5, Luma Ray 3.2, Kling 3 Pro, and Seedance 2.0. However, these figures stem from the early-access phase and do not constitute independent validation. Be that as it may, it's great to see the quality that today's video models can produce at affordable prices. Something that would have been unthinkable a year ago. Really cool release!

Black Forest Labs just announced FLUX 3: a unified multimodal model for image, video, audio and action prediction. Founded in Freiburg, Germany, one of the few globally significant European companies in the AI sector. The current release is gated early access, not open source. But if FLUX 3 Dev actually ships with usable open weights, this could become one of the most important open-model releases in generative video yet. Especially after the Chinese Minimax H3 release. BFL claims, based on preliminary internal comparisons, that FLUX 3 outperformed models such as Runway Gen-4.5, Luma Ray 3.2, Kling 3 Pro, and Seedance 2.0. However, these figures stem from the early-access phase and do not constitute independent validation. Be that as it may, it's great to see the quality that today's video models can produce at affordable prices. Something that would have been unthinkable a year ago. Really cool release!

39,869 просмотров

I'll always root for a team that open-sources its best work, and Robbyant just did it properly. Robbyant, Ant Group's embodied-AI company, released LingBot-Vision, a vision foundation model for robots, and the part I love is the data. They trained it on 161M images, filtered down from 2B raw ones and mostly pulled straight from the open web, with no human labels, no edge detectors, no depth sensors anywhere in the loop. It learns the exact edges of objects from raw pixels. That's roughly a tenth of the data DINOv3 saw, and under a third of the training. And it shows in the results. On depth, working out how far away things are, the 1B model edges out a 7B on NYU-Depth. It also powers LingBot-Depth 2.0, which reads the surfaces cameras usually choke on, glass and mirrors, and halves indoor depth error. LingBot-Vision is fully open. Weights from the 1.1B flagship down to a tiny 21M version, code, and the paper. This is the timeline I want more of. Robbyant

I'll always root for a team that open-sources its best work, and Robbyant just did it properly. Robbyant, Ant Group's embodied-AI company, released LingBot-Vision, a vision foundation model for robots, and the part I love is the data. They trained it on 161M images, filtered down from 2B raw ones and mostly pulled straight from the open web, with no human labels, no edge detectors, no depth sensors anywhere in the loop. It learns the exact edges of objects from raw pixels. That's roughly a tenth of the data DINOv3 saw, and under a third of the training. And it shows in the results. On depth, working out how far away things are, the 1B model edges out a 7B on NYU-Depth. It also powers LingBot-Depth 2.0, which reads the surfaces cameras usually choke on, glass and mirrors, and halves indoor depth error. LingBot-Vision is fully open. Weights from the 1.1B flagship down to a tiny 21M version, code, and the paper. This is the timeline I want more of. Robbyant

48,249 просмотров

Elon just said, "I think we are quite close to digital superintelligence. It may happen this year. If it doesn't happen this year, next year for sure. A digital superintelligence defined as smarter than any human at anything." Get ready, we are in for a wild ride.

Elon just said, "I think we are quite close to digital superintelligence. It may happen this year. If it doesn't happen this year, next year for sure. A digital superintelligence defined as smarter than any human at anything." Get ready, we are in for a wild ride.

286,250 просмотров

Demis Hassabis told world leaders: AGI will be 10x the impact of the industrial revolution at 10x the speed. So I used Perplexity computer to create a graph and visualized what it means for global GDP. No one is ready.

Demis Hassabis told world leaders: AGI will be 10x the impact of the industrial revolution at 10x the speed. So I used Perplexity computer to create a graph and visualized what it means for global GDP. No one is ready.

111,420 просмотров

/1 Gemma 4 31B just crushed Qwen 3.6 27B in a local LLM gamedev contest inside atomic.chat (prompt is below) Device: MacBook Pro M5 Max, 64GB RAM Results: Qwen 3.6 27B: 32 tokens/sec · 18m 04s · 33,946 tokens Gemma 4 31B: 27 tokens/sec · 3m 51s · 6,209 tokens So what is more important: tokens per second, or the quality of the final answer? Qwen made a very long response and showed more creativity and visual style. But Gemma gave a shorter, clearer, and more logical answer in much less time. In this one-shot Pac-Man gamedev contest, Gemma 4 31B was the clear winner. Its game logic was stronger: click reactions were smoother, and it handled interactions with elements like walls, ghosts, and particle effects better. But this was only one test. Maybe Qwen 3.6 27B can show better results with better settings. Open the comments, try our prompt, and share your result below.

/1 Gemma 4 31B just crushed Qwen 3.6 27B in a local LLM gamedev contest inside atomic.chat (prompt is below) Device: MacBook Pro M5 Max, 64GB RAM Results: Qwen 3.6 27B: 32 tokens/sec · 18m 04s · 33,946 tokens Gemma 4 31B: 27 tokens/sec · 3m 51s · 6,209 tokens So what is more important: tokens per second, or the quality of the final answer? Qwen made a very long response and showed more creativity and visual style. But Gemma gave a shorter, clearer, and more logical answer in much less time. In this one-shot Pac-Man gamedev contest, Gemma 4 31B was the clear winner. Its game logic was stronger: click reactions were smoother, and it handled interactions with elements like walls, ghosts, and particle effects better. But this was only one test. Maybe Qwen 3.6 27B can show better results with better settings. Open the comments, try our prompt, and share your result below.

72,566 просмотров

1/ Tell me again how blue-collar workers wont be replaced in the very near future. The robot Figure 02 had an 11-month deployment at BMW Group’s Spartanburg plant, including 10-hour daily shifts, during which it moved 90,000+ parts, ran 1,250+ hours, and contributed to the production of over 30,000 X3 vehicles. The deployment yielded critical learnings (especially about reliability and mechanical design) that are being incorporated into the next-gen robot Figure 03.

1/ Tell me again how blue-collar workers wont be replaced in the very near future. The robot Figure 02 had an 11-month deployment at BMW Group’s Spartanburg plant, including 10-hour daily shifts, during which it moved 90,000+ parts, ran 1,250+ hours, and contributed to the production of over 30,000 X3 vehicles. The deployment yielded critical learnings (especially about reliability and mechanical design) that are being incorporated into the next-gen robot Figure 03.

144,927 просмотров

1/ We've all been aware of the hype surrounding Dreamina Seedance 2.0, and I finally got early access to the tool. And wow, I'm blown away. This is by far the best. It is starting to make AI video feel less like "generate a clip" and more like "direct a scene." What stood out to me is the level of control: camera motion, pacing, visual consistency, and the ability to build from multiple references inside one workflow. Some of the prompt directions that feel especially strong: - a busy modern city square during daytime. Suddenly, time freezes completely - a single continuous camera movement through a natural landscape that transitions through all four seasons in one shot - an underwater bioluminescent city waking up at dawn The big shift is this: One Prompt, Viral Remade. Edit Videos as Easy as Editing Photos. Dreamina Seedance 2.0 feels like a real step toward AI-native directing rather than just AI generation. Here are some examples 🧵:

1/ We've all been aware of the hype surrounding Dreamina Seedance 2.0, and I finally got early access to the tool. And wow, I'm blown away. This is by far the best. It is starting to make AI video feel less like "generate a clip" and more like "direct a scene." What stood out to me is the level of control: camera motion, pacing, visual consistency, and the ability to build from multiple references inside one workflow. Some of the prompt directions that feel especially strong: - a busy modern city square during daytime. Suddenly, time freezes completely - a single continuous camera movement through a natural landscape that transitions through all four seasons in one shot - an underwater bioluminescent city waking up at dawn The big shift is this: One Prompt, Viral Remade. Edit Videos as Easy as Editing Photos. Dreamina Seedance 2.0 feels like a real step toward AI-native directing rather than just AI generation. Here are some examples 🧵:

76,342 просмотров

Waymo self-driving car enters active police standoff with passenger inside AGI achieved.

Waymo self-driving car enters active police standoff with passenger inside AGI achieved.

111,816 просмотров

Andrew Yang just returned from an AI conference, and the pace of innovation is staggering. Experts predict the next 6 months will outpace the last 10 years. The rate of change is on a hockey stick—and it blew his mind.

Andrew Yang just returned from an AI conference, and the pace of innovation is staggering. Experts predict the next 6 months will outpace the last 10 years. The rate of change is on a hockey stick—and it blew his mind.

70,210 просмотров

Videos

Anya Rossi

sweetdream.ai

SweetDream.ai•Sponsored•Livecam

Watch Anya Live

Anya is streaming live right now! Join her private show and enjoy exclusive content.

Exclusive private shows

1.2k viewers online

Private Show

Join now for exclusive access

Free preview available • Premium content

Anthropic’s Tom Brown: “My prediction is that within the next 12 months, AI could become a once-in-a-generation scientist in key scientific fields.” If you thought AI was slowing down or that we had already reached its peak, you couldn’t be more wrong. We barely started.

Anthropic’s Tom Brown: “My prediction is that within the next 12 months, AI could become a once-in-a-generation scientist in key scientific fields.” If you thought AI was slowing down or that we had already reached its peak, you couldn’t be more wrong. We barely started.

22,543 просмотров • 8 часов назад

AI video companies are no longer claiming they can simply generate better clips. They are claiming their models understand the world. So Peter and I put that claim to the test. We sat down with Yaron Inger, Co-Founder and CTO of LTX.io , for a direct conversation about LTX-2.5, its new open-weights video and world model. We asked what the model actually understands beyond producing plausible frames. Whether its physics are good enough for robotics. What native multi-shot generation really remembers across cuts. Whether its speed benchmarks are fair. What creators can expect on hardware they can realistically afford. And how LTX plans to build a business when the weights are available to download. This is not a product demo. It is a conversation about whether AI video is becoming genuine infrastructure for filmmaking, simulation, and physical AI. Watch the full interview:

AI video companies are no longer claiming they can simply generate better clips. They are claiming their models understand the world. So Peter and I put that claim to the test. We sat down with Yaron Inger, Co-Founder and CTO of LTX.io , for a direct conversation about LTX-2.5, its new open-weights video and world model. We asked what the model actually understands beyond producing plausible frames. Whether its physics are good enough for robotics. What native multi-shot generation really remembers across cuts. Whether its speed benchmarks are fair. What creators can expect on hardware they can realistically afford. And how LTX plans to build a business when the weights are available to download. This is not a product demo. It is a conversation about whether AI video is becoming genuine infrastructure for filmmaking, simulation, and physical AI. Watch the full interview:

17,742 просмотров • 9 часов назад

Another supposedly GPT-"Astra" output. One-shotted a GTA-1 clone using max reasoning. Unbelievably good, insane. h/t XIVIX

Another supposedly GPT-"Astra" output. One-shotted a GTA-1 clone using max reasoning. Unbelievably good, insane. h/t XIVIX

92,998 просмотров • 4 дней назад

Someone on Reddit built a WoW private server with 1,800 bots and AI chat via the DeepSeek API. Dead Internet Theory, but playable. An MMORPG with no real players, yet somehow it still feels human.

Someone on Reddit built a WoW private server with 1,800 bots and AI chat via the DeepSeek API. Dead Internet Theory, but playable. An MMORPG with no real players, yet somehow it still feels human.

2,597,257 просмотров • 2 месяцев назад

At this point, in dont get it anymore. Yes, it looks impressive. But whats the use-case other than proving that, well, it can get up fast and run?

At this point, in dont get it anymore. Yes, it looks impressive. But whats the use-case other than proving that, well, it can get up fast and run?

8,788,325 просмотров • 9 месяцев назад

This came as a surprise: Microsoft has unveiled handheld and desktop devices designed to control one's agents. It reminds me of what I had expected from OpenAI’s hardware-standalone devices for controlling agents.

This came as a surprise: Microsoft has unveiled handheld and desktop devices designed to control one's agents. It reminds me of what I had expected from OpenAI’s hardware-standalone devices for controlling agents.

1,293,593 просмотров • 3 месяцев назад

Tencent just released Hy4 Preview: a 770B-parameter open-source model with 49B active parameters and a 1M-token context window. The most interesting part is its agentic research capability. Tencent says Hy4 coordinated several Codex sessions in parallel, evaluated their results and adjusted the research direction, outperforming Codex working alone across eight benchmarks. Hy4 is aimed at long-horizon coding, multi-file office work and scientific research. In sum: a solid, good release. A few months ago this would be absolute sota.

Tencent just released Hy4 Preview: a 770B-parameter open-source model with 49B active parameters and a 1M-token context window. The most interesting part is its agentic research capability. Tencent says Hy4 coordinated several Codex sessions in parallel, evaluated their results and adjusted the research direction, outperforming Codex working alone across eight benchmarks. Hy4 is aimed at long-horizon coding, multi-file office work and scientific research. In sum: a solid, good release. A few months ago this would be absolute sota.

49,575 просмотров • 5 дней назад

Google cooked so hard. Not gonna lie, this feels like the future is here. Now develop Google Glasses with enough battery power, a good chip, and a look like Ray-Bans, and you'll have an instant hit. 100%.

Google cooked so hard. Not gonna lie, this feels like the future is here. Now develop Google Glasses with enough battery power, a good chip, and a look like Ray-Bans, and you'll have an instant hit. 100%.

3,111,084 просмотров • 9 месяцев назад

Billionaire and Scale AI CEO Alexandr Wang: DeepSeek has about 50,000 NVIDIA H100s that they can't talk about because of the US export controls that are in place.

Billionaire and Scale AI CEO Alexandr Wang: DeepSeek has about 50,000 NVIDIA H100s that they can't talk about because of the US export controls that are in place.

5,418,410 просмотров • 1 год назад

Tried Hy4 preview through WorkBuddy and gave it a pretty vague prompt for a first person roller coaster ride, front row seat, make it feel fast, that was basically it. No detailed spec, no reference video, just a rough idea in a sentence. What came back actually held up. Steep drop that felt like the bottom dropping out, banked turns that leaned the right way, scenery rushing by close enough that you register the speed instead of just watching it happen. For something built off a loose prompt like that, I did not expect it to nail the feel this well. Coding and 3D rendering have clearly gotten a real upgrade here, and this took one pass to look this good. Free to try in WorkBuddy for the next two weeks if you want to mess with it yourself. Tencent Hy Tencent AI

Tried Hy4 preview through WorkBuddy and gave it a pretty vague prompt for a first person roller coaster ride, front row seat, make it feel fast, that was basically it. No detailed spec, no reference video, just a rough idea in a sentence. What came back actually held up. Steep drop that felt like the bottom dropping out, banked turns that leaned the right way, scenery rushing by close enough that you register the speed instead of just watching it happen. For something built off a loose prompt like that, I did not expect it to nail the feel this well. Coding and 3D rendering have clearly gotten a real upgrade here, and this took one pass to look this good. Free to try in WorkBuddy for the next two weeks if you want to mess with it yourself. Tencent Hy Tencent AI

27,540 просмотров • 4 дней назад

Visualization of Convolutional Neural Network

Visualization of Convolutional Neural Network

3,848,609 просмотров • 1 год назад

Here we see GTA IV overlaid with AI generated graphics. I can very well imagine that in the near future we will generate graphics with AI, because it is cheaper and faster than if everything is created in advance in a graphics engine.

Here we see GTA IV overlaid with AI generated graphics. I can very well imagine that in the near future we will generate graphics with AI, because it is cheaper and faster than if everything is created in advance in a graphics engine.

4,064,773 просмотров • 1 год назад

Sam Altman: Superintelligence probably by end of 2028. So we got roughly 2 years left. Enjoy your job while you still can. Time is ticking.

Sam Altman: Superintelligence probably by end of 2028. So we got roughly 2 years left. Enjoy your job while you still can. Time is ticking.

1,069,214 просмотров • 6 месяцев назад

1/ Spent the past few days testing Lux3D, a 3D generative model. You give it an image or a text prompt and it hands you back a full, editable 3D asset. No modeling skills required. Tried all 3 core features. A thread:

1/ Spent the past few days testing Lux3D, a 3D generative model. You give it an image or a text prompt and it hands you back a full, editable 3D asset. No modeling skills required. Tried all 3 core features. A thread:

21,355 просмотров • 4 дней назад

People getting tricked by a fake AI influencer. Welcome to 2026.

People getting tricked by a fake AI influencer. Welcome to 2026.

1,100,164 просмотров • 7 месяцев назад

Walker S2 replacing it's own battery 24/7 blue-collar-worker within reach

Walker S2 replacing it's own battery 24/7 blue-collar-worker within reach

1,802,577 просмотров • 1 год назад

Nvidia showcases Blue, a cute little robot powered by the Newton physics engine. This was the „wow“-moment for me. Robots are coming. This time for real. And I am all in for it!

Nvidia showcases Blue, a cute little robot powered by the Newton physics engine. This was the „wow“-moment for me. Robots are coming. This time for real. And I am all in for it!

1,917,894 просмотров • 1 год назад

This is nuts; Elevenlabs nailed it. Voice but especially latency. After reading Matt Shumer's article, it's become even clearer to me what he means when he says that AI will soon encompass all other areas as well. Who needs call center agents when you have such a human-like AI?

This is nuts; Elevenlabs nailed it. Voice but especially latency. After reading Matt Shumer's article, it's become even clearer to me what he means when he says that AI will soon encompass all other areas as well. Who needs call center agents when you have such a human-like AI?

593,463 просмотров • 6 месяцев назад

If we manage to use the technology live in games without performance loss, that would be a huge step forward. RayTracing could be used without major performance losses, graphics could be made as realistic as possible and wanted and the style could be customized as desired. GenAI will also be used extensively in the gaming industry in the very near future!

If we manage to use the technology live in games without performance loss, that would be a huge step forward. RayTracing could be used without major performance losses, graphics could be made as realistic as possible and wanted and the style could be customized as desired. GenAI will also be used extensively in the gaming industry in the very near future!

1,993,082 просмотров • 1 год назад

I am seriously so annoyed by all those AI haters. DLSS5 looks phenomenal; people hate on it cause „it’s AI“. That’s the sole reason.

I am seriously so annoyed by all those AI haters. DLSS5 looks phenomenal; people hate on it cause „it’s AI“. That’s the sole reason.

484,319 просмотров • 5 месяцев назад