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。

GPT-6.1 Sol, running in Ultra mode, has been autonomously iterating on this project for more than 16 hours now. It’s using Blender and Three.js, and the entire thing runs directly in the browser. Like I said earlier, this model is genuinely impressive. OpenAI built something here that feels extremely... show more

Token Gremlin

10,622 subscribers

18,553 次观看 • 14 天前 •via X (Twitter)

教育游戏科学技术

Anya Rossi• Live Now

Private livecam show

29 条评论

Token Gremlin 的头像

Token Gremlin14 天前

@OpenAI @TheRealAdamG Fun fact: GPT-6 Astra recorded and edited the whole thing, and it took less than 5 minutes. I genuinely feel like I have a whole team of GPT Gremlins working for me.

Lucas Freire Sangoi 的头像

Lucas Freire Sangoi14 天前

@OpenAIDevs @OpenAI @TheRealAdamG fast enough? are you kidding me?

Jesus is God ✝︎ 的头像

Jesus is God ✝︎14 天前

@OpenAIDevs @OpenAI @TheRealAdamG

Meghraj 的头像

Meghraj14 天前

@OpenAIDevs @OpenAI @TheRealAdamG amazing.

Gregor 的头像

Gregor14 天前

@OpenAIDevs @OpenAI @TheRealAdamG 16 hours of autonomous tool calls is not a free demo

Isaiah 的头像

Isaiah14 天前

@OpenAIDevs @OpenAI @TheRealAdamG What plan is this on, this is insane

Tak 🦞 的头像

Tak 🦞14 天前

@OpenAIDevs @OpenAI @TheRealAdamG Cozy af

Pavel Moore 的头像

Pavel Moore14 天前

@OpenAIDevs @OpenAI @TheRealAdamG No wonder there servers are struggling lol. You are pushing it to the max

John Herzog 的头像

John Herzog14 天前

@OpenAIDevs @OpenAI @TheRealAdamG Sixteen hours of autonomous Blender and Three.js iteration in-browser is the first time long agent loops look cheap enough to keep running

✦ 𝓚𝓪𝓽𝓮𝓑𝓳𝓪𝓻𝓷𝓸𝓐𝓻𝓽 的头像

✦ 𝓚𝓪𝓽𝓮𝓑𝓳𝓪𝓻𝓷𝓸𝓐𝓻𝓽14 天前

@OpenAIDevs @OpenAI @TheRealAdamG That is very cool

Sola Awodiya 的头像

Sola Awodiya14 天前

@OpenAIDevs @OpenAI @TheRealAdamG Crazy

John Herzog 的头像

John Herzog14 天前

@OpenAIDevs @OpenAI @TheRealAdamG under five minutes from record to cut is when the “team of GPT Gremlins” line stops sounding like a joke

Memedei 的头像

Memedei14 天前

@OpenAIDevs @OpenAI @TheRealAdamG Whats about limits ?

The SI Therapist 的头像

The SI Therapist14 天前

@OpenAIDevs @OpenAI @TheRealAdamG The AI builds the app while I nap. My kind of work day. Hope it runs smooth in Chrome

Brian Spaeth 的头像

Brian Spaeth14 天前

@OpenAIDevs @OpenAI @TheRealAdamG I built this - 5.6 Sol actually did most of it but 6.1 has been huge in getting it finished

nmz 的头像

nmz14 天前

@OpenAIDevs @OpenAI @TheRealAdamG 2012-level garbage.

Vishal 的头像

Vishal14 天前

@OpenAIDevs @OpenAI @TheRealAdamG 16h is wild, i get nervous leaving mine alone for 2h without peeking at the diff. you checking in on it or just letting it cook?

Progressing Tech 的头像

Progressing Tech14 天前

@OpenAIDevs @OpenAI @TheRealAdamG How much of your weekly usage has it taken up?

James Malsawm 的头像

James Malsawm14 天前

@OpenAIDevs @OpenAI @TheRealAdamG It's a good results. What about the speed of making this?

安叫兽|Bird🕊️ 🔶 BNB 的头像

安叫兽|Bird🕊️ 🔶 BNB13 天前

@OpenAIDevs @OpenAI @TheRealAdamG 跑了十六个小时，想看看中途都改了些啥

EDDY VU 的头像

EDDY VU14 天前

@OpenAIDevs @OpenAI @TheRealAdamG Did it manage to recover from runtime errors cleanly on its own over that entire 16 hour stretch?

Haleemah 的头像

Haleemah13 天前

@OpenAIDevs @OpenAI @TheRealAdamG How long did this take to complete ?

🇺🇦🩸🤍 @cutest_star1🌟🩸🇺🇦 的头像

🇺🇦🩸🤍 @cutest_star1🌟🩸🇺🇦13 天前

@OpenAIDevs @OpenAI @TheRealAdamG Looks retarded and devoid of any merit or interesting aspects

Ishu Agrawal 的头像

Ishu Agrawal14 天前

@OpenAIDevs @OpenAI @TheRealAdamG great, now do Breath of the Wild 😂

sheilu 的头像

sheilu14 天前

@OpenAIDevs @OpenAI @TheRealAdamG How much

Raghav Katta 的头像

Raghav Katta14 天前

@OpenAIDevs @OpenAI @TheRealAdamG for a long-running browser scene build, how are you tracking whether each iteration improves the result? a fixed camera route plus frame-time and visual checkpoints could help distinguish useful progress from a scene that is simply becoming more complicated.

Paolo 的头像

Paolo14 天前

@OpenAIDevs @OpenAI @TheRealAdamG were u not a Plus user? Ultramode?

Claudio 的头像

Claudio13 天前

@OpenAIDevs @OpenAI @TheRealAdamG How are you managing this usage in Ultra on a Plus plan? Or are you on a Pro plan?

Bharath Gedela 的头像

Bharath Gedela14 天前

@OpenAIDevs @OpenAI @TheRealAdamG How does it choose what to improve after the first playable version—are you giving it a checklist, or is it reviewing the result itself?

相关视频

⚡️OPENAI: GPT-5.6 SOL IS 54% MORE TOKEN EFFICIENT ON AGENTIC CODING Sam Altman says the new GPT-5.6 Sol model is 54% more token efficient on agentic coding tasks and is “as good or better” than competing models. He also said OpenAI aims to offer better value with “every enterprise now is thinking about spend and the value they’re getting in exchange for AI.”

⚡️OPENAI: GPT-5.6 SOL IS 54% MORE TOKEN EFFICIENT ON AGENTIC CODING Sam Altman says the new GPT-5.6 Sol model is 54% more token efficient on agentic coding tasks and is “as good or better” than competing models. He also said OpenAI aims to offer better value with “every enterprise now is thinking about spend and the value they’re getting in exchange for AI.”

Coin Bureau

33,775 次观看 • 3 个月前

📣 OpenAI Developers GPT-6.1 Sol is now generally available in GitHub Copilot. You can use it for agentic coding and terminal workflows with strong multistep coding performance. In early testing, it reliably completed tasks while using noticeably fewer tokens and steps than earlier models in the GPT-6 and GPT-5.6 families. Try it out in the GitHub Copilot app, CLI, and Visual Studio Code.

📣 OpenAI Developers GPT-6.1 Sol is now generally available in GitHub Copilot. You can use it for agentic coding and terminal workflows with strong multistep coding performance. In early testing, it reliably completed tasks while using noticeably fewer tokens and steps than earlier models in the GPT-6 and GPT-5.6 families. Try it out in the GitHub Copilot app, CLI, and Visual Studio Code.

GitHub

118,609 次观看 • 15 天前

One thing I did which has been extremely cool is have Grok 4.5 in Grok Build CLI scan my project for all the UI/UX bugs it could find and make a detailed list. Got a couple hundred. Then had GPT 5.6 Sol in Cursor fix all of them by delegating subagents and managing the PRs. It’s been running for 3 days now and handled an insane number of them, and is still going. I’ve been checking in every couple hours and it’s just extremely good work.

One thing I did which has been extremely cool is have Grok 4.5 in Grok Build CLI scan my project for all the UI/UX bugs it could find and make a detailed list. Got a couple hundred. Then had GPT 5.6 Sol in Cursor fix all of them by delegating subagents and managing the PRs. It’s been running for 3 days now and handled an insane number of them, and is still going. I’ve been checking in every couple hours and it’s just extremely good work.

Max Weinbach

46,570 次观看 • 3 个月前

This is The Ramada Inn Project in Los Angeles It’s a homeless housing project that was supposed to only cost $4.8 million for the renovation 6 years later that renovation is now $21 million and the only thing fixed for $21 million was new windows…. Money laundering I looked more into this and apparently of as of right now in 2026, construction and renovation is finally underway, with an expected opening in late 2026 Clearly this is a lie. Since the property still looks horrible and nothing but windows has been done The cost is $640,000 per unit for just 32 apartments. This is extremely high even by LA standards This project has been sitting unfinished for so long it’s not even a Karen Bass project, it’s the previous mayor Eric Garcetti Multiple terms of Democrat after Democrat failure Vote Spencer Pratt

This is The Ramada Inn Project in Los Angeles It’s a homeless housing project that was supposed to only cost $4.8 million for the renovation 6 years later that renovation is now $21 million and the only thing fixed for $21 million was new windows…. Money laundering I looked more into this and apparently of as of right now in 2026, construction and renovation is finally underway, with an expected opening in late 2026 Clearly this is a lie. Since the property still looks horrible and nothing but windows has been done The cost is $640,000 per unit for just 32 apartments. This is extremely high even by LA standards This project has been sitting unfinished for so long it’s not even a Karen Bass project, it’s the previous mayor Eric Garcetti Multiple terms of Democrat after Democrat failure Vote Spencer Pratt

Wall Street Apes

180,386 次观看 • 4 个月前

GPT-6.1 Sol ULTRA ran for 25 minutes and only used 1% of my weekly quota, It’s seriously fast and powerful, almost like Astra, but much cheaper to run, For me, this is the kind of model you can actually use 24/7 without constantly worrying about limits, From now on, it’s probably going to be my default model, Honestly, out of everything announced at DevDay, this might be my favorite update. In my experience, it feels like the most efficient model OpenAI has released so far

GPT-6.1 Sol ULTRA ran for 25 minutes and only used 1% of my weekly quota, It’s seriously fast and powerful, almost like Astra, but much cheaper to run, For me, this is the kind of model you can actually use 24/7 without constantly worrying about limits, From now on, it’s probably going to be my default model, Honestly, out of everything announced at DevDay, this might be my favorite update. In my experience, it feels like the most efficient model OpenAI has released so far

Iam_

141,848 次观看 • 15 天前

Welcome to GPT-6.1 Sol ultrafast! Here is a small demo of me making very quick changes in a native SwiftUI iOS app using ultrafast. It's is up to 8X faster than standard mode and open many workflows with the right agentic setup. 🧵

Welcome to GPT-6.1 Sol ultrafast! Here is a small demo of me making very quick changes in a native SwiftUI iOS app using ultrafast. It's is up to 8X faster than standard mode and open many workflows with the right agentic setup. 🧵

Thomas Ricouard

68,863 次观看 • 6 天前

GPT-6.1 Sol isn’t that good I asked it to build the Eiffel Tower in Three.js, and this is what it came up with. It performs pretty well on some other tests, but it failed badly on this scene compared to other models. For me, it still feels below Astra and even Kimi in frontend and 3D. OpenAI says it’s more efficient, but it seems to consume a lot of tokens.

GPT-6.1 Sol isn’t that good I asked it to build the Eiffel Tower in Three.js, and this is what it came up with. It performs pretty well on some other tests, but it failed badly on this scene compared to other models. For me, it still feels below Astra and even Kimi in frontend and 3D. OpenAI says it’s more efficient, but it seems to consume a lot of tokens.

Bhavy☄️

38,564 次观看 • 15 天前

$GPT 6.1 Sol is cooking asked model to add skateboarding to GTA V and this is what it made in 4 hours it’s still working, but all of this was done with 6.1 sol max and barely used any limits team really cooked harder with this model we now have a model with astra level intelligence at a fraction of the cost$

GPT 6.1 Sol is cooking asked model to add skateboarding to GTA V and this is what it made in 4 hours it’s still working, but all of this was done with 6.1 sol max and barely used any limits team really cooked harder with this model we now have a model with astra level intelligence at a fraction of the cost

J A Z I I

250,863 次观看 • 14 天前

At this point, Three.js is probably the best graphics stack for AI-generated games. Not because it magically produces better visuals, but because models are extremely good at reasoning directly over JavaScript, scene graphs, shaders, procedural geometry, physics integrations, and fast browser iteration. The examples here make the case pretty clearly. Once multiple agents start building scenes in isolation and integrating them later, the ceiling gets surprisingly high. For AI-native game development, Three.js currently feels like the path of least resistance to genuinely impressive results. I gathered a few videos from Chetan in this demo and combined them.

At this point, Three.js is probably the best graphics stack for AI-generated games. Not because it magically produces better visuals, but because models are extremely good at reasoning directly over JavaScript, scene graphs, shaders, procedural geometry, physics integrations, and fast browser iteration. The examples here make the case pretty clearly. Once multiple agents start building scenes in isolation and integrating them later, the ceiling gets surprisingly high. For AI-native game development, Three.js currently feels like the path of least resistance to genuinely impressive results. I gathered a few videos from Chetan in this demo and combined them.

Token Gremlin

12,038 次观看 • 2 个月前

GPT-5.6 Sol Ultra was given a full set of character references and asked to rebuild the character in Blender, rig it, create three animations, export a real GLB, and build a complete Three.js showcase around it. The entire pipeline was completed in 2 hours and 11 minutes. The final model still loses some of the original proportions, costume structure, and material fidelity, but going from reference images to a rigged, animated 3D character running in the browser is already a very strong result. Cc: Givros

GPT-5.6 Sol Ultra was given a full set of character references and asked to rebuild the character in Blender, rig it, create three animations, export a real GLB, and build a complete Three.js showcase around it. The entire pipeline was completed in 2 hours and 11 minutes. The final model still loses some of the original proportions, costume structure, and material fidelity, but going from reference images to a rigged, animated 3D character running in the browser is already a very strong result. Cc: Givros

Token Gremlin

37,583 次观看 • 3 个月前

Opus 5.5 made a full Splatoon Inspired Game. Here is Version 3.0 with the new map. (First clip) This is the version that will be shipped with multiplayer. Everything has been revamped and built from the ground up. The game now has procedural animations for character movement and dynamic paint splatter. The game also features Xbox Controller support and vibrations. This was made ENTIRELY built in two days and it is running in the browser. Powered by Three.js

Opus 5.5 made a full Splatoon Inspired Game. Here is Version 3.0 with the new map. (First clip) This is the version that will be shipped with multiplayer. Everything has been revamped and built from the ground up. The game now has procedural animations for character movement and dynamic paint splatter. The game also features Xbox Controller support and vibrations. This was made ENTIRELY built in two days and it is running in the browser. Powered by Three.js

Jayden Davis

16,547 次观看 • 19 天前

I got Black Ops 2 running natively in Minecraft. Not through a browser, not on a screen inside the game. I ported Hijacked to Java and it runs directly on Minecraft’s own OpenGL context. The map, collision, bots, navmesh, weapons and rendering are all running inside Minecraft as a Fabric mod. This started with me putting my browser port on a PC screen inside Minecraft. That worked, but I wanted to see if I could go much further and actually port the game itself. So I had GPT Astra rebuild the whole thing in Java. Still a lot left to do, but this is now an actual game running inside Minecraft rather than Minecraft displaying one.

I got Black Ops 2 running natively in Minecraft. Not through a browser, not on a screen inside the game. I ported Hijacked to Java and it runs directly on Minecraft’s own OpenGL context. The map, collision, bots, navmesh, weapons and rendering are all running inside Minecraft as a Fabric mod. This started with me putting my browser port on a PC screen inside Minecraft. That worked, but I wanted to see if I could go much further and actually port the game itself. So I had GPT Astra rebuild the whole thing in Java. Still a lot left to do, but this is now an actual game running inside Minecraft rather than Minecraft displaying one.

Luckey Faraday

880,525 次观看 • 28 天前

Infinite GPT-6 for everybody. GPT-6 is finally in chatgpt and chat has no limit so you just TURN CHATGPT INTO CODEX with Chat on Steroids. Agents, skills, loop, goal, plan mode, terminal, computer use, everything codex has. Meanwhile codex limits are dogshit and 6.1 sol is slow as fuck. Tibo said sorry for it and one week later the fast 6.1 sol is 500$ pro only lol. Video is GPT-6 high fixing a broken project by itself in like a minute and a half

Infinite GPT-6 for everybody. GPT-6 is finally in chatgpt and chat has no limit so you just TURN CHATGPT INTO CODEX with Chat on Steroids. Agents, skills, loop, goal, plan mode, terminal, computer use, everything codex has. Meanwhile codex limits are dogshit and 6.1 sol is slow as fuck. Tibo said sorry for it and one week later the fast 6.1 sol is 500$ pro only lol. Video is GPT-6 high fixing a broken project by itself in like a minute and a half

Binkg

28,400 次观看 • 5 天前

Can't believe in less than 24 hours I created my own game with Claude and Three.js. Running in a browser at 120 FPS. All I had to do was give Claude the repo for Threejs and describe an MMO-like game for it to build from scratch. All the assets and mechanics you see are all from a prompt. As you can see, it is far from done but the progress is insane. Everyone should be using Three.js now without a doubt!

Can't believe in less than 24 hours I created my own game with Claude and Three.js. Running in a browser at 120 FPS. All I had to do was give Claude the repo for Threejs and describe an MMO-like game for it to build from scratch. All the assets and mechanics you see are all from a prompt. As you can see, it is far from done but the progress is insane. Everyone should be using Three.js now without a doubt!

NeilXbt

23,504 次观看 • 2 个月前

This is MiniMax-M2.5 MLX running in LM Studio on an Apple Mac Studio M3 Ultra 512GB. Fast enough out of the box for hosting OpenClaw, n8n workflows, and Open WebUI for the team.

This is MiniMax-M2.5 MLX running in LM Studio on an Apple Mac Studio M3 Ultra 512GB. Fast enough out of the box for hosting OpenClaw, n8n workflows, and Open WebUI for the team.

Patrick J Kennedy

76,702 次观看 • 8 个月前

gpt-5.5 just dropped. and it changes EVERYTHING for traders. openai stopped calling it a chat model. they call it an "agent runtime" now. that means one thing: AI that doesn't wait for you - it acts. here's what gpt-5.5 can actually do: → runs multi-step trading workflows autonomously → operates computer interfaces without human input → plans, iterates, executes - in a terminal, not a chat window → 82.7% on Terminal-Bench 2.0 - real command-line task execution the only question is: are you already running one? full FREE Guide here → like + bookmark. this is the wave.

gpt-5.5 just dropped. and it changes EVERYTHING for traders. openai stopped calling it a chat model. they call it an "agent runtime" now. that means one thing: AI that doesn't wait for you - it acts. here's what gpt-5.5 can actually do: → runs multi-step trading workflows autonomously → operates computer interfaces without human input → plans, iterates, executes - in a terminal, not a chat window → 82.7% on Terminal-Bench 2.0 - real command-line task execution the only question is: are you already running one? full FREE Guide here → like + bookmark. this is the wave.

AlphaCartel

56,823 次观看 • 5 个月前

This is probably my stopping point right here, as I've been working on this for the last four to five hours, and the costs of spending more time on this are outweighing the benefits. The issue I'm running into now is that it doesn't seem to be agentic enough to be able to do stuff on its own that is meaningful, such as build apps generatively and take my feedback. I do think that the technology has a lot of promise, so I will check back in a couple weeks or months to try and pick up where I left off.

This is probably my stopping point right here, as I've been working on this for the last four to five hours, and the costs of spending more time on this are outweighing the benefits. The issue I'm running into now is that it doesn't seem to be agentic enough to be able to do stuff on its own that is meaningful, such as build apps generatively and take my feedback. I do think that the technology has a lot of promise, so I will check back in a couple weeks or months to try and pick up where I left off.

Pat Walls

44,125 次观看 • 8 个月前

Anthropic is in trouble and nobody wants to say it out loud. Opus 4.7 is the first Claude model people think is worse, usage limits are hitting fast, compute is running thin, and Anthropic's best model still isn't out. GPT-5.5 just dropped today, OpenAI found their window. This 32 minute video is worth more than the next episode of whatever you're watching on Netflix tonight.

Anthropic is in trouble and nobody wants to say it out loud. Opus 4.7 is the first Claude model people think is worse, usage limits are hitting fast, compute is running thin, and Anthropic's best model still isn't out. GPT-5.5 just dropped today, OpenAI found their window. This 32 minute video is worth more than the next episode of whatever you're watching on Netflix tonight.

Defileo🔮

131,838 次观看 • 5 个月前

Andrew Ng, on the biggest shift in AI right now: “The most important trend is agentic AI workflows.” In one slide, he showed the part most people still miss: GPT-3.5 scored 48% on a coding benchmark, GPT-4 reached 67%, but GPT-3.5 inside an agentic workflow jumped to around 95%. That is not a prompt trick. It is the model planning, using tools, checking its own work, fixing mistakes and running the loop again. Everyone is still arguing about which model is smarter, while the real edge is moving to the system wrapped around the model. Watch the clip, then build the workflow around your own process below.

Andrew Ng, on the biggest shift in AI right now: “The most important trend is agentic AI workflows.” In one slide, he showed the part most people still miss: GPT-3.5 scored 48% on a coding benchmark, GPT-4 reached 67%, but GPT-3.5 inside an agentic workflow jumped to around 95%. That is not a prompt trick. It is the model planning, using tools, checking its own work, fixing mistakes and running the loop again. Everyone is still arguing about which model is smarter, while the real edge is moving to the system wrapped around the model. Watch the clip, then build the workflow around your own process below.

Gipp 🦅

17,267 次观看 • 3 个月前

🚨 STOP learning Blender !! Learning Blender in 2026..... Is officially obsolete. MiniMax Design can now automate Blender to create ANY 3d SCENE !! I just gave it a PROMPT.... And MiniMax Design used GPT-6 Astra to handle everything like: 1️⃣ Opening Blender 2️⃣ Generating the 3D elements 3️⃣ Fixing lighting 4️⃣ Camera Tracking Data etc.... It built the entire scene inside Blender. Finally, I used MiniMax H3 Max to add colors to the 3d SCENE !! Then I just rendered the 3d scene !! Best part? Using GPT-6 Astra through MiniMax Design is way more affordable than running it directly.

🚨 STOP learning Blender !! Learning Blender in 2026..... Is officially obsolete. MiniMax Design can now automate Blender to create ANY 3d SCENE !! I just gave it a PROMPT.... And MiniMax Design used GPT-6 Astra to handle everything like: 1️⃣ Opening Blender 2️⃣ Generating the 3D elements 3️⃣ Fixing lighting 4️⃣ Camera Tracking Data etc.... It built the entire scene inside Blender. Finally, I used MiniMax H3 Max to add colors to the 3d SCENE !! Then I just rendered the 3d scene !! Best part? Using GPT-6 Astra through MiniMax Design is way more affordable than running it directly.

Darshal Jaitwar

86,087 次观看 • 1 个月前