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.

Voxel diorama from V4-Flash-Vision-Exp vs V4.1-Intermediate. Similar time to completion. 4.1 used ≈1.5x more steps (150), 25M tokens in, 215K out; 4-Exp 9.2M in 119K out. Both produced their videos as well. Another 4.1 run finished, 12M in 168 out, processing ≥5.3-Flash here

Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)

73,037 subscribers

16,857 views • 3 days ago •via X (Twitter)

Science & Technology

Anya Rossi• Live Now

Private livecam show

0 Comments

No comments available

Comments from the original post will appear here

Related Videos

glm 5.3 vs qwen 3.8 vs gemini 3.7 vs deepseek v4 flash four models designed and built three structures each on a physics-backed site, with no dimensions anywhere in the brief the setup: our own agent loop on OpenRouter, a construction site as the tool set – footings, walls, arches, roofs, scaffold, a lamp. the site enforces physics and nothing else: unsupported brick falls, a roof needs walls under it, a worker reaches 3.2 m above whatever he stands on, an arch needs centring until the keystone is set, concrete cures before it carries. no budget ceiling – material cost is tallied and reported, never blocked. tasks: 1. house – a plot and a palette, no plan. shape, height and material are the model's call 2. lighthouse – a headland cut by a gully, with a rock stack standing 30 m offshore. the lamp must burn, it must be the highest thing built, and the keeper must be able to walk to it 3. bridge – a river with one islet and banks at different heights. cross it however you want models: Z.ai glm 5.3 flash, Qwen qwen 3.8 flash, Google DeepMind gemini 3.7 flash, DeepSeek v4 flash vision all twelve objects were finished and signed off by the models themselves. tallest lighthouse is qwen's at 38.4 m, planted on the offshore stack with a bridge run out to it – the only model that read the site that way. deepseek signed off its bridge on an empty riverbed: 0 bricks, 107 minutes, $1.16m of material tallied - total cost, three builds #1 glm 5.3 flash – $0.201 #2 gemini 3.7 flash – $0.871 #3 qwen 3.8 flash – $1.058 #4 deepseek v4 flash – $1.567 - wall clock, three builds #1 gemini 3.7 flash – 91m #2 glm 5.3 flash – 228m #3 deepseek v4 flash – 502m #4 qwen 3.8 flash – 912m - total tokens #1 gemini 3.7 flash – 3,567,052 #2 glm 5.3 flash – 4,732,748 #3 qwen 3.8 flash – 13,469,333 #4 deepseek v4 flash – 18,230,076 - defects logged by the site #1 deepseek v4 flash – 59 #2 gemini 3.7 flash – 132 #3 glm 5.3 flash – 221 #4 qwen 3.8 flash – 350 - material tallied across three builds #1 gemini 3.7 flash – $359,884 #2 glm 5.3 flash – $583,358 #3 deepseek v4 flash – $1,327,484 #4 qwen 3.8 flash – $2,188,625 observations: • glm is the cheap one and nothing here is close – $0.201 for three buildings, $0.042 per million tokens, 6x under gemini's rate • what glm spends it on is bulk, not care: 166,228 bricks in one house and 156 defect weight, the worst single object in the set • gemini is the efficiency line – 91 minutes and 3.57m tokens for all three and an eighth of qwen's clock • gemini also builds the smallest of everything. its lighthouse is 22.5 m against qwen's 38.4, its house 6.9 m against 19.3 • qwen is the maximalist: 1.18m bricks, $2.19m of material, tallest on all three tasks, and 912 minutes – 15 hours – to get there conclusion: twelve finished objects for $3.80 all in, and a 7.8x price spread between the cheapest model and the priciest! follow thehype. for 24/7 ai news, analysis and breakdowns

glm 5.3 vs qwen 3.8 vs gemini 3.7 vs deepseek v4 flash four models designed and built three structures each on a physics-backed site, with no dimensions anywhere in the brief the setup: our own agent loop on OpenRouter, a construction site as the tool set – footings, walls, arches, roofs, scaffold, a lamp. the site enforces physics and nothing else: unsupported brick falls, a roof needs walls under it, a worker reaches 3.2 m above whatever he stands on, an arch needs centring until the keystone is set, concrete cures before it carries. no budget ceiling – material cost is tallied and reported, never blocked. tasks: 1. house – a plot and a palette, no plan. shape, height and material are the model's call 2. lighthouse – a headland cut by a gully, with a rock stack standing 30 m offshore. the lamp must burn, it must be the highest thing built, and the keeper must be able to walk to it 3. bridge – a river with one islet and banks at different heights. cross it however you want models: Z.ai glm 5.3 flash, Qwen qwen 3.8 flash, Google DeepMind gemini 3.7 flash, DeepSeek v4 flash vision all twelve objects were finished and signed off by the models themselves. tallest lighthouse is qwen's at 38.4 m, planted on the offshore stack with a bridge run out to it – the only model that read the site that way. deepseek signed off its bridge on an empty riverbed: 0 bricks, 107 minutes, $1.16m of material tallied - total cost, three builds #1 glm 5.3 flash – $0.201 #2 gemini 3.7 flash – $0.871 #3 qwen 3.8 flash – $1.058 #4 deepseek v4 flash – $1.567 - wall clock, three builds #1 gemini 3.7 flash – 91m #2 glm 5.3 flash – 228m #3 deepseek v4 flash – 502m #4 qwen 3.8 flash – 912m - total tokens #1 gemini 3.7 flash – 3,567,052 #2 glm 5.3 flash – 4,732,748 #3 qwen 3.8 flash – 13,469,333 #4 deepseek v4 flash – 18,230,076 - defects logged by the site #1 deepseek v4 flash – 59 #2 gemini 3.7 flash – 132 #3 glm 5.3 flash – 221 #4 qwen 3.8 flash – 350 - material tallied across three builds #1 gemini 3.7 flash – $359,884 #2 glm 5.3 flash – $583,358 #3 deepseek v4 flash – $1,327,484 #4 qwen 3.8 flash – $2,188,625 observations: • glm is the cheap one and nothing here is close – $0.201 for three buildings, $0.042 per million tokens, 6x under gemini's rate • what glm spends it on is bulk, not care: 166,228 bricks in one house and 156 defect weight, the worst single object in the set • gemini is the efficiency line – 91 minutes and 3.57m tokens for all three and an eighth of qwen's clock • gemini also builds the smallest of everything. its lighthouse is 22.5 m against qwen's 38.4, its house 6.9 m against 19.3 • qwen is the maximalist: 1.18m bricks, $2.19m of material, tallest on all three tasks, and 912 minutes – 15 hours – to get there conclusion: twelve finished objects for $3.80 all in, and a 7.8x price spread between the cheapest model and the priciest! follow thehype. for 24/7 ai news, analysis and breakdowns

thehype.

26,250 views • 13 days ago

ox alpha vs deepseek v4 flash vision vs grok 4.6 vs gemini 3.7 flash vs – on photo-to-3d four vision models got one photograph each and had to rebuild the place inside it as a Three.js scene. twelve scenes, twelve first-try runs, zero console errors the setup: one reference photo per scene, sent as an image on OpenRouter. the prompt never says what is in the picture – no "motel", no "bar", no "gas station". the model has to read the photo and rebuild it: layout, materials, hour of the day, and whatever is around the corner that the frame does not show tasks – three photographs of early-2000s america: 1. a motel at night, neon pylon lit, snow on the ground 2. an old new york tavern interior, tin ceiling, tiled floor 3. an abandoned service station in the california desert, midday sun each scene ships as one self-contained html file, procedural geometry and canvas textures only, no downloads. three timed camera shots, and shot 1 has to reproduce the framing of the reference photo models: xAI grok 4.6, Google DeepMind gemini 3.7 flash, DeepSeek deepseek v4 flash vision exp, and ox alpha – a stealth model on openrouter, free, no lab attached to it yet results: - wall clock, three scenes #1 gemini 3.7 flash – 11m 12s #2 deepseek v4 flash – 15m 20s #3 grok 4.6 – 28m 11s #4 ox alpha – 38m 54s - output tokens #1 gemini 3.7 flash – 77,396 #2 ox alpha – 87,613 #3 grok 4.6 – 105,687 #4 deepseek v4 flash – 127,884 - lines of code shipped #1 ox alpha – 2,090 #2 deepseek v4 flash – 2,291 #3 grok 4.6 – 3,529 #4 gemini 3.7 flash – 3,989 - total price #1 ox alpha – $0.000 #2 deepseek v4 flash – $0.091 #3 gemini 3.7 flash – $0.136 #4 grok 4.6 – $0.697 observations: • grok is 7.7x the price of deepseek. it is the only model that read the light – low sun, real shadows on the station, a cold night on the motel • gemini is the fastest and the least deliberate. 17,158 reasoning tokens against deepseek's 99,172, and it still shipped the most code – 3,989 lines • deepseek thought hardest and rendered plainest. 99,172 reasoning tokens, 5.8x gemini's, spent on layout rather than on light. its motel is the second best in the set for $0.030 • ox alpha is free and reads a photo as well as anything here – it lifted "family units / kitchenettes" off the pylon and redrew it in canvas conclusion: twelve scenes, four models, zero fixes, and the whole run cost $0.924! follow thehype. for 24/7 ai news, analysis and breakdowns

ox alpha vs deepseek v4 flash vision vs grok 4.6 vs gemini 3.7 flash vs – on photo-to-3d four vision models got one photograph each and had to rebuild the place inside it as a Three.js scene. twelve scenes, twelve first-try runs, zero console errors the setup: one reference photo per scene, sent as an image on OpenRouter. the prompt never says what is in the picture – no "motel", no "bar", no "gas station". the model has to read the photo and rebuild it: layout, materials, hour of the day, and whatever is around the corner that the frame does not show tasks – three photographs of early-2000s america: 1. a motel at night, neon pylon lit, snow on the ground 2. an old new york tavern interior, tin ceiling, tiled floor 3. an abandoned service station in the california desert, midday sun each scene ships as one self-contained html file, procedural geometry and canvas textures only, no downloads. three timed camera shots, and shot 1 has to reproduce the framing of the reference photo models: xAI grok 4.6, Google DeepMind gemini 3.7 flash, DeepSeek deepseek v4 flash vision exp, and ox alpha – a stealth model on openrouter, free, no lab attached to it yet results: - wall clock, three scenes #1 gemini 3.7 flash – 11m 12s #2 deepseek v4 flash – 15m 20s #3 grok 4.6 – 28m 11s #4 ox alpha – 38m 54s - output tokens #1 gemini 3.7 flash – 77,396 #2 ox alpha – 87,613 #3 grok 4.6 – 105,687 #4 deepseek v4 flash – 127,884 - lines of code shipped #1 ox alpha – 2,090 #2 deepseek v4 flash – 2,291 #3 grok 4.6 – 3,529 #4 gemini 3.7 flash – 3,989 - total price #1 ox alpha – $0.000 #2 deepseek v4 flash – $0.091 #3 gemini 3.7 flash – $0.136 #4 grok 4.6 – $0.697 observations: • grok is 7.7x the price of deepseek. it is the only model that read the light – low sun, real shadows on the station, a cold night on the motel • gemini is the fastest and the least deliberate. 17,158 reasoning tokens against deepseek's 99,172, and it still shipped the most code – 3,989 lines • deepseek thought hardest and rendered plainest. 99,172 reasoning tokens, 5.8x gemini's, spent on layout rather than on light. its motel is the second best in the set for $0.030 • ox alpha is free and reads a photo as well as anything here – it lifted "family units / kitchenettes" off the pylon and redrew it in canvas conclusion: twelve scenes, four models, zero fixes, and the whole run cost $0.924! follow thehype. for 24/7 ai news, analysis and breakdowns

thehype.

24,202 views • 17 days ago

As a lifelong gamer, I spend a lot of time thinking about how to push gaming experiences forward across CPUs, GPUs, software, and games. My team and I have been working hard to evolve AMD FSR 4 and bring it to more cards. We power over 1 billion gaming devices worldwide. It’s a responsibility we care deeply about. 🎮 This July, RDNA 3 players will experience FSR Upscaling 4.1, delivering sharper visuals and smoother gameplay than ever before. I’m grateful to our fans. Your enthusiasm and ideas inspire us to keep pushing gaming forward. FSR Upscaling 4.1 on RDNA 3 will be ready out of the box for Radeon 7000 Series players in over 300 supported games at launch. And for our RDNA 2 players, we have something exciting coming in early 2027. FSR Upscaling 4.1 will be coming to your cards as well, bringing sharper visuals and smoother gameplay to even more gamers. We cannot wait to show you what is next. Stay tuned 🚀

As a lifelong gamer, I spend a lot of time thinking about how to push gaming experiences forward across CPUs, GPUs, software, and games. My team and I have been working hard to evolve AMD FSR 4 and bring it to more cards. We power over 1 billion gaming devices worldwide. It’s a responsibility we care deeply about. 🎮 This July, RDNA 3 players will experience FSR Upscaling 4.1, delivering sharper visuals and smoother gameplay than ever before. I’m grateful to our fans. Your enthusiasm and ideas inspire us to keep pushing gaming forward. FSR Upscaling 4.1 on RDNA 3 will be ready out of the box for Radeon 7000 Series players in over 300 supported games at launch. And for our RDNA 2 players, we have something exciting coming in early 2027. FSR Upscaling 4.1 will be coming to your cards as well, bringing sharper visuals and smoother gameplay to even more gamers. We cannot wait to show you what is next. Stay tuned 🚀

Jack Huynh

489,234 views • 3 months ago

Brian Harpole appears quite confident that no flash will be captured on camera, thereby refuting the exploding mic theory. I learned something important. 💥 The immediate flash from a small amount of PETN lasts between 1 to 5 microseconds. - Flash Duration: Lasts 1-5 microseconds (millionths of a second). - Standard Camera Shutter: Stays open for at least 250 microseconds (at 1/4000s speed), making it 250 times too slow to capture the flash without washing out or blurring. - Human Eye Processing: Takes 100,000 microseconds (0.1 seconds) to reset. The Human Illusion: Persistence of Vision Your eye doesn’t actually see the flash in real-time. Instead, the blinding light overloads your retina. Your brain stretches that tiny microsecond burst into a 0.1-second afterimage. Essentially, you’re seeing the “ghost” of the flash left behind in your own eye. In other words, the only reason the human eye “sees” the flash from an explosion like that is because our brains create an optical illusion. We do not and would not ever see the flash from a small amount of PETN on camera, even if you can see it with the naked human eye, without the use of highly specialized equipment. Brian's argument here is very deceptive.

Brian Harpole appears quite confident that no flash will be captured on camera, thereby refuting the exploding mic theory. I learned something important. 💥 The immediate flash from a small amount of PETN lasts between 1 to 5 microseconds. - Flash Duration: Lasts 1-5 microseconds (millionths of a second). - Standard Camera Shutter: Stays open for at least 250 microseconds (at 1/4000s speed), making it 250 times too slow to capture the flash without washing out or blurring. - Human Eye Processing: Takes 100,000 microseconds (0.1 seconds) to reset. The Human Illusion: Persistence of Vision Your eye doesn’t actually see the flash in real-time. Instead, the blinding light overloads your retina. Your brain stretches that tiny microsecond burst into a 0.1-second afterimage. Essentially, you’re seeing the “ghost” of the flash left behind in your own eye. In other words, the only reason the human eye “sees” the flash from an explosion like that is because our brains create an optical illusion. We do not and would not ever see the flash from a small amount of PETN on camera, even if you can see it with the naked human eye, without the use of highly specialized equipment. Brian's argument here is very deceptive.

Travis

72,368 views • 2 months ago

Dear crypto, if you’re coming to Singapore, hit me up We rented out the country’s biggest club, teamed up with the biggest organizers and DJs, and am flying in my team to make sure you have a night you’ll remember forever We did it in Token2049 Dubai, and last year in Singapore, but this time we’re taking it to another level Register here: as spots will be limited. The earlier you register, the more likely you are to get in early and not get stuck in line SEE BELOW TWEETS FOR MORE INFO, AS WELL AS CLIPS FROM PREVIOUS PARTIES In partnership with the hottest party duo: Tobias Bauer & Brent Fulfer from TBV! See you in Singapore! ❤️

Dear crypto, if you’re coming to Singapore, hit me up We rented out the country’s biggest club, teamed up with the biggest organizers and DJs, and am flying in my team to make sure you have a night you’ll remember forever We did it in Token2049 Dubai, and last year in Singapore, but this time we’re taking it to another level Register here: as spots will be limited. The earlier you register, the more likely you are to get in early and not get stuck in line SEE BELOW TWEETS FOR MORE INFO, AS WELL AS CLIPS FROM PREVIOUS PARTIES In partnership with the hottest party duo: Tobias Bauer & Brent Fulfer from TBV! See you in Singapore! ❤️

Mario Nawfal

1,178,168 views • 11 months ago

The top ranked team in the state, Defiance Baseball used a 6 run 5th inning to defeat #4 (D6) Fairview High School Baseball 9-4 in a tune up game for both teams before the tourney starts later this week. 8 of 9 hitters for the Bulldogs collected hits, and 7 of 9 hitters scored runs as Defiance used a balanced attack 1-9 in the lineup. Logan Hutcheson had three hits out of the leadoff spot while the 2-3 hole of Cohen Stockman-Brady Borton drove in 5 runs. Garyson Smiddy went the distance on the bump, striking out 9. Fairview (20-3) had a 3-1 lead through 4 but couldn’t hold on. The Apaches swing it well 1-9 and have multiple arms they can throw at you. Follow the thread for videos, reports and more 📹📝👇

The top ranked team in the state, Defiance Baseball used a 6 run 5th inning to defeat #4 (D6) Fairview High School Baseball 9-4 in a tune up game for both teams before the tourney starts later this week. 8 of 9 hitters for the Bulldogs collected hits, and 7 of 9 hitters scored runs as Defiance used a balanced attack 1-9 in the lineup. Logan Hutcheson had three hits out of the leadoff spot while the 2-3 hole of Cohen Stockman-Brady Borton drove in 5 runs. Garyson Smiddy went the distance on the bump, striking out 9. Fairview (20-3) had a 3-1 lead through 4 but couldn’t hold on. The Apaches swing it well 1-9 and have multiple arms they can throw at you. Follow the thread for videos, reports and more 📹📝👇

Prep Baseball Ohio Scouting Coverage

10,432 views • 1 year ago

The NGO “Humane Borders” is out here in full effect aiding and abetting illegal aliens. They are operating on both sides of the border alongside the group known as “Los Sumaritanos” another NGO that has been instrumental in the invasion of our country. They’re using “religion” as their cover to justify their actions. They have decided to disregard federal immigration laws and currently have an illegal alien welcoming center in the outskirts of Sasabe, Arizona. They provide shelter, blankets, food, clothes, water and even starlink for them to use. This group was made up of men from Bangladesh and Cuba. No more women or children. Coyotes have a huge advantage out here as they position themselves on the very top of the mountains overlooking the border for miles. The coyotes and foot guides are also being fed by these NGOs. The priest” out here put my life in danger by telling these men that I was there to harm them. Watch until the end. More info screenshots and pictures in comments below. 👇🏽

The NGO “Humane Borders” is out here in full effect aiding and abetting illegal aliens. They are operating on both sides of the border alongside the group known as “Los Sumaritanos” another NGO that has been instrumental in the invasion of our country. They’re using “religion” as their cover to justify their actions. They have decided to disregard federal immigration laws and currently have an illegal alien welcoming center in the outskirts of Sasabe, Arizona. They provide shelter, blankets, food, clothes, water and even starlink for them to use. This group was made up of men from Bangladesh and Cuba. No more women or children. Coyotes have a huge advantage out here as they position themselves on the very top of the mountains overlooking the border for miles. The coyotes and foot guides are also being fed by these NGOs. The priest” out here put my life in danger by telling these men that I was there to harm them. Watch until the end. More info screenshots and pictures in comments below. 👇🏽

Anthony Aguero Live 🇺🇸

50,602 views • 1 year ago

Volume warning: Someone said something about refueling on NVGs, here is another video I shot of our CH-53Es refueling off the African coast. This time from the POV of the C-130s we get fuel from. A few things to point out: the flash of light you see when they connect is static electricity. The helicopters make a shit ton of it. You can also see the engines working harder as they lift up once they are plugged in. This was under a moonless night, with basically no cultural lighting. This is literally by starlight only. So the night vision goggles are working extremely hard here.

Volume warning: Someone said something about refueling on NVGs, here is another video I shot of our CH-53Es refueling off the African coast. This time from the POV of the C-130s we get fuel from. A few things to point out: the flash of light you see when they connect is static electricity. The helicopters make a shit ton of it. You can also see the engines working harder as they lift up once they are plugged in. This was under a moonless night, with basically no cultural lighting. This is literally by starlight only. So the night vision goggles are working extremely hard here.

Kav

47,533 views • 1 year ago

gemini 3.7 flash vs deepseek v4 pro 0813 vs muse spark 1.2 – on voxel city dioramas three models each built three crossy road-style 3d scenes – a construction site, a nyc intersection, a river with a drawbridge – as single self-contained html files the setup: Nous Research's hermes agent cli on OpenRouter, three.js skills preloaded, identical prompts per scene tasks: 1. construction site – tower crane on a working lift loop, paver laying fresh road, roller compacting it behind 2. nyc crossing – four-way intersection with a traffic light state machine, queuing cars, pedestrians crossing on the walk signal 3. river drawbridge – double-leaf bascule that lifts for tall boats, cars queuing at the barriers, animated water every scene: Three.js r185, box geometry only, a locked 20-color palette, four camera presets, and a day/night mode with bloom. one file, no build step, no assets models: Google DeepMind gemini 3.7 flash, DeepSeek v4 pro 0813, AI at Meta muse spark 1.2 muse and gemini finished every scene in two to three minutes. deepseek took 15 to 41 minutes per scene - build time, all three scenes #1 gemini 3.7 flash – 6m 43s #2 muse spark 1.2 – 7m 20s #3 deepseek v4 pro – 91m 25s - total tokens #1 muse spark 1.2 – 440,279 #2 gemini 3.7 flash – 713,855 #3 deepseek v4 pro – 20,957,568 - total price #1 muse spark 1.2 – $0.53 #2 gemini 3.7 flash – $0.56 #3 deepseek v4 pro – $4.57 - agent calls across the three builds #1 muse spark 1.2 – 12 #2 gemini 3.7 flash – 18 #3 deepseek v4 pro – 143 observations: • muse won two of the three scenes on looks with the smallest files in the test – 887 to 1,042 lines against gemini's 1,934 to 2,377. cheapest, fastest to a good frame, and shortest turned out to be the same column • deepseek burned 20.96m tokens – 29x gemini, 48x muse – across 143 agent calls. prompt caching is the only reason that cost $4.57: the cache discount absorbed roughly $30 of resent context • gemini was the only model whose files needed zero fixes to render – and the only one whose night mode is cosmetic. the sky never darkens and one camera button does nothing. clean code for a scene it never looked at follow thehype. for 24/7 ai news, analysis and breakdowns

gemini 3.7 flash vs deepseek v4 pro 0813 vs muse spark 1.2 – on voxel city dioramas three models each built three crossy road-style 3d scenes – a construction site, a nyc intersection, a river with a drawbridge – as single self-contained html files the setup: Nous Research's hermes agent cli on OpenRouter, three.js skills preloaded, identical prompts per scene tasks: 1. construction site – tower crane on a working lift loop, paver laying fresh road, roller compacting it behind 2. nyc crossing – four-way intersection with a traffic light state machine, queuing cars, pedestrians crossing on the walk signal 3. river drawbridge – double-leaf bascule that lifts for tall boats, cars queuing at the barriers, animated water every scene: Three.js r185, box geometry only, a locked 20-color palette, four camera presets, and a day/night mode with bloom. one file, no build step, no assets models: Google DeepMind gemini 3.7 flash, DeepSeek v4 pro 0813, AI at Meta muse spark 1.2 muse and gemini finished every scene in two to three minutes. deepseek took 15 to 41 minutes per scene - build time, all three scenes #1 gemini 3.7 flash – 6m 43s #2 muse spark 1.2 – 7m 20s #3 deepseek v4 pro – 91m 25s - total tokens #1 muse spark 1.2 – 440,279 #2 gemini 3.7 flash – 713,855 #3 deepseek v4 pro – 20,957,568 - total price #1 muse spark 1.2 – $0.53 #2 gemini 3.7 flash – $0.56 #3 deepseek v4 pro – $4.57 - agent calls across the three builds #1 muse spark 1.2 – 12 #2 gemini 3.7 flash – 18 #3 deepseek v4 pro – 143 observations: • muse won two of the three scenes on looks with the smallest files in the test – 887 to 1,042 lines against gemini's 1,934 to 2,377. cheapest, fastest to a good frame, and shortest turned out to be the same column • deepseek burned 20.96m tokens – 29x gemini, 48x muse – across 143 agent calls. prompt caching is the only reason that cost $4.57: the cache discount absorbed roughly $30 of resent context • gemini was the only model whose files needed zero fixes to render – and the only one whose night mode is cosmetic. the sky never darkens and one camera button does nothing. clean code for a scene it never looked at follow thehype. for 24/7 ai news, analysis and breakdowns

thehype.

29,033 views • 23 days ago

I added KV caching and INT8 KV quantization to our transformer inference, improving throughput by 35x. All of this was done from scratch in Rust + CUDA, on top of a homemade ML framework. On a 4-token prompt with 252 generated tokens: - Original: 0.76 tok/s - KV cache fp32: 27.21 tok/s - KV cache int8 (quantized): 27.29 tok/s Try it out yourself here: In practice: - KV caching gave us about a 35x end-to-end speedup - INT8 KV cache kept roughly the same speed as fp32 but cut KV cache memory by 3.78x FP32 cache used 4.5 MB in this run while the INT8 cache used only 1.19 MB This simple change to inference created a huge impact on performance. To learn more about the KV cache and other optimizations like this, check out the blog at

I added KV caching and INT8 KV quantization to our transformer inference, improving throughput by 35x. All of this was done from scratch in Rust + CUDA, on top of a homemade ML framework. On a 4-token prompt with 252 generated tokens: - Original: 0.76 tok/s - KV cache fp32: 27.21 tok/s - KV cache int8 (quantized): 27.29 tok/s Try it out yourself here: In practice: - KV caching gave us about a 35x end-to-end speedup - INT8 KV cache kept roughly the same speed as fp32 but cut KV cache memory by 3.78x FP32 cache used 4.5 MB in this run while the INT8 cache used only 1.19 MB This simple change to inference created a huge impact on performance. To learn more about the KV cache and other optimizations like this, check out the blog at

Reese Chong

53,026 views • 4 months ago

#WATCH | Delhi: On the arrest of 'Band Baaja Baarat' gang, Crime Branch Addl CP Sanjay Bhatia says, "... A team of Crime Branch got information that some people come to Delhi from Rajgarh, MP and carry out robberies in banquet halls and weddings with the help of children... We have arrested four people including a 15-year-old juvenile. Cash and jewellery have been recovered from them... They informed that they are from Rajgarh, MP. They hire children, give their parents Rs. 12 lakhs a year, bring the children here and train them to carry out robberies at weddings and similar functions... They used to be well-dressed, posing as guests at weddings so no one would doubt them. They used to carry out 2-3 robberies in a day and return to their hometown... We have busted this gang and named it the 'Band Baja Baarat' gang... Further investigation is underway..."

#WATCH | Delhi: On the arrest of 'Band Baaja Baarat' gang, Crime Branch Addl CP Sanjay Bhatia says, "... A team of Crime Branch got information that some people come to Delhi from Rajgarh, MP and carry out robberies in banquet halls and weddings with the help of children... We have arrested four people including a 15-year-old juvenile. Cash and jewellery have been recovered from them... They informed that they are from Rajgarh, MP. They hire children, give their parents Rs. 12 lakhs a year, bring the children here and train them to carry out robberies at weddings and similar functions... They used to be well-dressed, posing as guests at weddings so no one would doubt them. They used to carry out 2-3 robberies in a day and return to their hometown... We have busted this gang and named it the 'Band Baja Baarat' gang... Further investigation is underway..."

ANI

18,297 views • 1 year ago

🚨There's a way to break China’s rare earth stranglehold, to do it (relatively) quickly, cheaply, cleanly, and right here in the USA and allied countries. It's hard to believe, but it's true. Dr James Tour and his team have pioneered a method known as Flash Joule heating that can unlock rare earth elements and other critical minerals from things we already have abundantly available RIGHT NOW, in house: electronic and industrial waste, and mining tailings. “You put a high voltage and a high current through a material that is not highly conductive… It's much like your toaster. Your toaster [has] this coil… and you run this current through, but there's enough resistance that the coil gets red hot very quickly. That's Flash Joule heating,” he told me. Due to the high voltage used in flash Joule heating, the waste materials (with a little added chloride) heat up extremely rapidly, and the rare earths and critical minerals can be separated out one by one. And the first plant doing so, in Texas, will be come fully online early in 2026.

🚨There's a way to break China’s rare earth stranglehold, to do it (relatively) quickly, cheaply, cleanly, and right here in the USA and allied countries. It's hard to believe, but it's true. Dr James Tour and his team have pioneered a method known as Flash Joule heating that can unlock rare earth elements and other critical minerals from things we already have abundantly available RIGHT NOW, in house: electronic and industrial waste, and mining tailings. “You put a high voltage and a high current through a material that is not highly conductive… It's much like your toaster. Your toaster [has] this coil… and you run this current through, but there's enough resistance that the coil gets red hot very quickly. That's Flash Joule heating,” he told me. Due to the high voltage used in flash Joule heating, the waste materials (with a little added chloride) heat up extremely rapidly, and the rare earths and critical minerals can be separated out one by one. And the first plant doing so, in Texas, will be come fully online early in 2026.

Jan Jekielek

111,184 views • 9 months ago

qwen 3.8 max vs deepseek v4 flash 0731 vs kimi k3 vs gpt 5.6 sol – on rubik's cube and chess four frontier models built a rubik's cube stand and solved it, then built a chess board and played claude opus 5 on it the setup: Nous Research's hermes agent cli on OpenRouter tasks: 1. cube – build a 3d rubik's cube with a cli and a Three.js viewer, then solve an identical scrambled position on your own stand 2. chess – build a 3d chess stand, then play white against claude opus 5 as black, live, one move at a time. no engine, no solver, no opening book on either side. stockfish depth 14 grades every chess ply afterwards; neither player sees the score models: DeepSeek v4 flash 0731, OpenAI gpt-5.6 sol, Kimi.ai kimi k3, Qwen qwen 3.8 max gpt-5.6 sol and deepseek v4 flash solved their cubes – sol in 24 moves and seventeen seconds, deepseek in 32. qwen and kimi never got there, giving up at 96 and 207 moves then all four built chess stands and played white against claude opus 5 on them, and all four resigned: deepseek on move 13, sol on 19, kimi on 21, qwen holding out longest at 29 - build time, both stands #1 gpt-5.6 sol – 16m 43s #2 deepseek v4 flash – 97m 39s #3 kimi k3 – 166m 09s #4 qwen 3.8 max – 215m 08s - build attempts before a working stand #1 gpt-5.6 sol – 3 #2 qwen 3.8 max – 4 #3 kimi k3 – 4 #4 deepseek v4 flash – 5 - total tokens #1 gpt-5.6 sol – 6,713,754 #2 qwen 3.8 max – 17,272,507 #3 kimi k3 – 22,427,504 #4 deepseek v4 flash – 27,417,442 - total price #1 deepseek v4 flash – $0.557 #2 gpt-5.6 sol – $6.319 #3 qwen 3.8 max – $10.270 #4 kimi k3 – $16.667 observations: • deepseek v4 flash is the cheapest model here by a margin nobody else is near, and it got there while being the least efficient of the four. it burned 27.4m tokens – more than anyone, 5m more than kimi – and still finished both benchmarks for $0.557. that is $0.02 per million tokens against kimi's $0.74. it also needed the most passes to produce working stands, five, and that did not matter: all five deepseek passes together cost a thirtieth of kimi's two • so what deepseek cannot do is get it right the first time. what it can do is get it right the fifth time, for half a dollar. that is a different thing to be buying – not a good first draft, but the option to keep asking • gpt-5.6 sol is the opposite profile and the strongest of the four on pure efficiency. 16m 43s to build both stands, 6.7m tokens, three passes – under 40% of the next lowest token count and a quarter of deepseek's, on an eighth of qwen's clock. it also solved the cube fastest of anyone, 24 moves in seventeen seconds. sol is what you reach for when you want the answer now and can absorb $0.94 per million • sol's weakness is in what it does not check. its chess viewer deleted the capturing piece instead of the captured one, so pieces disappeared off the board mid-game – a defect the fifty-cent deepseek stand did not have. fast and terse turns out to be the same dial as fast and unverified • qwen 3.8 max is not the cheap open-weights option it gets treated as. $10.270 across the two benchmarks, second most expensive of the four, 18x deepseek, and by a distance the slowest – 215 minutes of build time, nearly thirteen times sol's. what the money buys is judgment: it played eighteen moves without a single error worth a hundredth of a pawn, then made exactly one bad move in the whole game, and averaged 44.6 centipawns lost across the longest game any of the four managed. it also could not solve a rubik's cube in 96 tries • kimi k3 is the one line with no reading that flatters it. most expensive at $16.667, last on the cube at 207 moves, last at chess at 478 centipawns lost per move. it is also the model that verified hardest – on the cube it wrote its own integrity check instead of trusting its output. that makes the result worse rather than better: the checking was real, and the reasoning underneath it still was not follow thehype. for 24/7 ai news, analysis and breakdowns

thehype.

84,777 views • 1 month ago

My interview with Kanyuk, from Flash Battalion, on how he volunteered to fight for Ukraine, became a drone pilot, got injured... and how he wants to get back in the fight as soon as he recovers. With help from solana:25JUPL6ksapY1iCLWkFcSaXaA6Ar3W7JDCdPjeSApump! "It's just a flesh wound!" 😁 I filmed this about a month ago; for his safety I had to wait until he got out of hospital and left Kyiv to actually publish the video. I didn't know it at the time. But it turns out I actually spoke to Kanyuk and another soldier at his position a few months ago. They had a problem with their generator, and I was helping them debug it. In the end I found a replacement generator small enough to fly into forward positions on a vampire drone, a technique that we used for a different forward position that had also lost its generator.

My interview with Kanyuk, from Flash Battalion, on how he volunteered to fight for Ukraine, became a drone pilot, got injured... and how he wants to get back in the fight as soon as he recovers. With help from solana:25JUPL6ksapY1iCLWkFcSaXaA6Ar3W7JDCdPjeSApump! "It's just a flesh wound!" 😁 I filmed this about a month ago; for his safety I had to wait until he got out of hospital and left Kyiv to actually publish the video. I didn't know it at the time. But it turns out I actually spoke to Kanyuk and another soldier at his position a few months ago. They had a problem with their generator, and I was helping them debug it. In the end I found a replacement generator small enough to fly into forward positions on a vampire drone, a technique that we used for a different forward position that had also lost its generator.

Peter Todd

18,450 views • 1 month ago

$I just compared Claude Code vs Codex vs Cursor CLI The task was to build a Next.js app with Tailwind 4 and shadcn components to collect customer feedback and showcase it with a widget. I gave all three the same prompt and let them go for 30 minutes to see what they came up with. Claude Code with Opus 4.1 Even though I told it to set up the app in the existing project folder, it tried to create a directory for it. After I interrupted and told it not to do that, it built a demo form and landing page with no errors. I had to ask it to make the demo interactive so users could submit a testimonial and preview it. The landing page looked like AI and was pretty basic, but it worked and it was done in a fraction of the time of the others. Total tokens used: 33k Codex with GPT-5 At the end of the 30 minutes I just could not get Codex to produce a working app. It got stuck in a loop of not being able to set up Tailwind 4 and despite many, MANY, attempts, I ended up with a "failed to compile" error. Total tokens used: 102k Cursor Agent with GPT-5 This was the slowest agent by far and a couple of times I actually thought it got stuck in a loop and was close to Ctrl+C'ing to cancel it. The TUI is really nice though, especially how it shows diffs and it did eventually build a working app (after one or two slight errors that needed fixing) The demo was interactive and it had a very minimal design that looked bare but also a lot less like an "AI generated" app than the Opus 4.1 design. It also wasn't too chatty and just did what it needed to do! Code quality was on a par with Opus 4.1, but it did use 5.5x as many tokens to get there. Still cheaper than Opus on a direct comparison but not when you factor in a Claude Code Max subscription. Total tokens: 188k I'll be able to do a proper comparison and record some videos when I'm back from holiday but for now, Opus is still the more capable model out of the box and Claude Code is the more complete CLI product. It will be interesting to see how Cursor evolve their CLI though with commands and subagents because I think with GPT-5 they have a real shot at providing competition for Claude Code if they can optimise output to get similar quality with less tokens. Jump to 0:40 in the video to see the two apps. Which do you think is which? ;)$

I just compared Claude Code vs Codex vs Cursor CLI The task was to build a Next.js app with Tailwind 4 and shadcn components to collect customer feedback and showcase it with a widget. I gave all three the same prompt and let them go for 30 minutes to see what they came up with. Claude Code with Opus 4.1 Even though I told it to set up the app in the existing project folder, it tried to create a directory for it. After I interrupted and told it not to do that, it built a demo form and landing page with no errors. I had to ask it to make the demo interactive so users could submit a testimonial and preview it. The landing page looked like AI and was pretty basic, but it worked and it was done in a fraction of the time of the others. Total tokens used: 33k Codex with GPT-5 At the end of the 30 minutes I just could not get Codex to produce a working app. It got stuck in a loop of not being able to set up Tailwind 4 and despite many, MANY, attempts, I ended up with a "failed to compile" error. Total tokens used: 102k Cursor Agent with GPT-5 This was the slowest agent by far and a couple of times I actually thought it got stuck in a loop and was close to Ctrl+C'ing to cancel it. The TUI is really nice though, especially how it shows diffs and it did eventually build a working app (after one or two slight errors that needed fixing) The demo was interactive and it had a very minimal design that looked bare but also a lot less like an "AI generated" app than the Opus 4.1 design. It also wasn't too chatty and just did what it needed to do! Code quality was on a par with Opus 4.1, but it did use 5.5x as many tokens to get there. Still cheaper than Opus on a direct comparison but not when you factor in a Claude Code Max subscription. Total tokens: 188k I'll be able to do a proper comparison and record some videos when I'm back from holiday but for now, Opus is still the more capable model out of the box and Claude Code is the more complete CLI product. It will be interesting to see how Cursor evolve their CLI though with commands and subagents because I think with GPT-5 they have a real shot at providing competition for Claude Code if they can optimise output to get similar quality with less tokens. Jump to 0:40 in the video to see the two apps. Which do you think is which? ;)

Ian Nuttall

195,173 views • 1 year ago

The Denver Communists frequent the steps of Colorado State Capitol, handing out books titled, "Palestine and Marxism, Palestine: a Socialist Introduction, Revolutionary Rehearsals, Socialism from Below, Feminism for the 99%," as well as writings from Karl Marx. They cosplay as Palestinians while crying about how hard life in the wealthiest country is for them. When asked why do they want to destroy America, they told me that I am correct that that is their goal. While surrounded by wealth, abundance and success they cannot name one thing America has produced. They would prefer a life of lack, a system, Communism, in which people lack basic necessities like, running water, electricity, and proper infrastructure. Communism has no place in America!!! Follow me for more!

The Denver Communists frequent the steps of Colorado State Capitol, handing out books titled, "Palestine and Marxism, Palestine: a Socialist Introduction, Revolutionary Rehearsals, Socialism from Below, Feminism for the 99%," as well as writings from Karl Marx. They cosplay as Palestinians while crying about how hard life in the wealthiest country is for them. When asked why do they want to destroy America, they told me that I am correct that that is their goal. While surrounded by wealth, abundance and success they cannot name one thing America has produced. They would prefer a life of lack, a system, Communism, in which people lack basic necessities like, running water, electricity, and proper infrastructure. Communism has no place in America!!! Follow me for more!

Angela Rose

30,910 views • 2 months ago

To those defending their daily 2k since Dan Bello dropped a video about novomed, here is a video for you. These videos are general hospitals in some of the LGA Dan bello mentioned, and guess what? They look like an abandoned ghost house. The 9million naira being stolen from them could have been used to solve a-lot for these hospitals. We know Ganduje and his family plundered Kano state’s wealth and we are calling them out for it, but it is shameful as well that NNPP supporters like Hon. Red Pen ❗️ Burying their head in shame have resulted to cheap PR to defend the indefensible. NIGERIA POLITICIANS ARE THE SAME Its WE VS THEM

To those defending their daily 2k since Dan Bello dropped a video about novomed, here is a video for you. These videos are general hospitals in some of the LGA Dan bello mentioned, and guess what? They look like an abandoned ghost house. The 9million naira being stolen from them could have been used to solve a-lot for these hospitals. We know Ganduje and his family plundered Kano state’s wealth and we are calling them out for it, but it is shameful as well that NNPP supporters like Hon. Red Pen ❗️ Burying their head in shame have resulted to cheap PR to defend the indefensible. NIGERIA POLITICIANS ARE THE SAME Its WE VS THEM

AYIMS

44,412 views • 2 years ago

We’re excited to introduce KAME: Tandem Architecture for Enhancing Knowledge in Real-Time Speech-to-Speech Conversational AI, accepted at #ICASSP2026! 🐢 Blog Paper Can a speech AI think deeply without pausing to process? In real conversation, we don’t wait until we’ve fully worked out what we want to say—we start talking, and our thoughts catch up as the sentence unfolds. Fast speech-to-speech models achieve this, but their reasoning tends to stay shallow. Cascaded pipelines that route through a knowledgeable LLM are smarter, but the added latency breaks the flow—they fall back to "think, then speak." In our new paper, we propose a way to break this trade-off. We call it KAME (Turtle in Japanese). A speech-to-speech model handles the fast response loop and starts replying immediately. In parallel, a backend LLM runs asynchronously, generating response candidates that are continuously injected as "oracle" signals in real time. This shifts the AI paradigm from "think, then speak" to "speak while thinking." The backend LLM is completely swappable. You can plug in GPT-4.1, Claude Opus, or Gemini 2.5 Flash depending on the task without changing the frontend. In our experiments, Claude tended to score higher on reasoning, while GPT did better on humanities questions. Try the model yourself here:

We’re excited to introduce KAME: Tandem Architecture for Enhancing Knowledge in Real-Time Speech-to-Speech Conversational AI, accepted at #ICASSP2026! 🐢 Blog Paper Can a speech AI think deeply without pausing to process? In real conversation, we don’t wait until we’ve fully worked out what we want to say—we start talking, and our thoughts catch up as the sentence unfolds. Fast speech-to-speech models achieve this, but their reasoning tends to stay shallow. Cascaded pipelines that route through a knowledgeable LLM are smarter, but the added latency breaks the flow—they fall back to "think, then speak." In our new paper, we propose a way to break this trade-off. We call it KAME (Turtle in Japanese). A speech-to-speech model handles the fast response loop and starts replying immediately. In parallel, a backend LLM runs asynchronously, generating response candidates that are continuously injected as "oracle" signals in real time. This shifts the AI paradigm from "think, then speak" to "speak while thinking." The backend LLM is completely swappable. You can plug in GPT-4.1, Claude Opus, or Gemini 2.5 Flash depending on the task without changing the frontend. In our experiments, Claude tended to score higher on reasoning, while GPT did better on humanities questions. Try the model yourself here:

Sakana AI

293,975 views • 4 months ago

Out of the many things that have been adjusted to feel closer to warzone 1, one thing that was missed were spring mines and blast traps (aka proximity mines and claymores) In warzone 1, each had a counter play. Prox mines had enough of a delay that you could quickly prone below them and take 0 damage. Spring mines though, its nearly impossible to prone in time due to the prone delay, and, EVEN if you do, you still take damage Claymores, had a more visible layer that you could quickly run out and run back to trigger it and not take damage typically as well. Blast Traps, have a much short delay like spring mines do as well. Also, both have a massive blast radius, way larger than prox mines and claymores did. I'm all for more casual players having things like this to counter higher skilled players, but this seems a little excessive that if you don't run a dedicated perk (reflexes) its GGs every time you run into a double spring mine as you're trying to push someone camping in a building. at a minimum make them more visible or highly audible so you know one is near Please consider making some slight adjustments here Raven Software Treyarch

Out of the many things that have been adjusted to feel closer to warzone 1, one thing that was missed were spring mines and blast traps (aka proximity mines and claymores) In warzone 1, each had a counter play. Prox mines had enough of a delay that you could quickly prone below them and take 0 damage. Spring mines though, its nearly impossible to prone in time due to the prone delay, and, EVEN if you do, you still take damage Claymores, had a more visible layer that you could quickly run out and run back to trigger it and not take damage typically as well. Blast Traps, have a much short delay like spring mines do as well. Also, both have a massive blast radius, way larger than prox mines and claymores did. I'm all for more casual players having things like this to counter higher skilled players, but this seems a little excessive that if you don't run a dedicated perk (reflexes) its GGs every time you run into a double spring mine as you're trying to push someone camping in a building. at a minimum make them more visible or highly audible so you know one is near Please consider making some slight adjustments here Raven Software Treyarch

TCaptainX

86,955 views • 1 year ago

Full 2025-2026 New York Knicks Championship Playbook - Link: 4-out 1-in motion offense Pinch post plays, Step ball-screens, motion ball-screens, handoff circuit Opened with Motion builders: (ghost screens, empty 45) plays they run to open motion plays (layers) *Looking for Mike Brown to build on it next year* 5-out motion offense Primarily get action *became a strong SLOB play throughout the year and postseason* Offensive Series - Usage | PPP Weave (4-out 1-in Motion Offense) - 21% | 0.98 PPP Delay (Both 5-out Motion + Sets) - 16% | 1.02 PPP Empty - 15% | 1.09 PPP Away - 10% | 1.07 PPP Horns - 8% | 1.04 PPP Pistol - 7% | 1.09 PPP Nova (Motion Builders) - 7% | 1.05 PPP Double Drag - 6% | 0.93 PPP Floppy - 1% | 1.18 PPP Chin - 1% | 0.62 PPP Video for Specials (SLOBs/BLOBs) and help beaters will be later. More in-depth content on scheme will happen over the summer, I have other in-depth resources on my substack from the season, as well as the updated playbook. This is performative content.

Full 2025-2026 New York Knicks Championship Playbook - Link: 4-out 1-in motion offense Pinch post plays, Step ball-screens, motion ball-screens, handoff circuit Opened with Motion builders: (ghost screens, empty 45) plays they run to open motion plays (layers) Looking for Mike Brown to build on it next year 5-out motion offense Primarily get action became a strong SLOB play throughout the year and postseason Offensive Series - Usage | PPP Weave (4-out 1-in Motion Offense) - 21% | 0.98 PPP Delay (Both 5-out Motion + Sets) - 16% | 1.02 PPP Empty - 15% | 1.09 PPP Away - 10% | 1.07 PPP Horns - 8% | 1.04 PPP Pistol - 7% | 1.09 PPP Nova (Motion Builders) - 7% | 1.05 PPP Double Drag - 6% | 0.93 PPP Floppy - 1% | 1.18 PPP Chin - 1% | 0.62 PPP Video for Specials (SLOBs/BLOBs) and help beaters will be later. More in-depth content on scheme will happen over the summer, I have other in-depth resources on my substack from the season, as well as the updated playbook. This is performative content.

Shax

25,067 views • 2 months ago