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.

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

Chubby♨️

139,129 subscribers

93,531 views • 4 days ago •via X (Twitter)

Science & Technology Gaming

Anya Rossi• Live Now

Private livecam show

0 Comments

No comments available

Comments from the original post will appear here

Related Videos

Another WTF moment. Using Codex as a fully custom game engine With the newly discovered and *insane* img2threejs skill 1) GPT imageGen a 3D item you want 2) ask GPT 5.6 Sol to use img2threejs 3) my harness brings it into the game automatically Just watch how good this is

Another WTF moment. Using Codex as a fully custom game engine With the newly discovered and insane img2threejs skill 1) GPT imageGen a 3D item you want 2) ask GPT 5.6 Sol to use img2threejs 3) my harness brings it into the game automatically Just watch how good this is

Nicolas Zullo

267,097 views • 1 month ago

I told Claude 4.6 Opus to make a pokemon clone - max effort It reasoned for 1 hour and 30 minutes and used 110k tokens and 2 shotted this absolute behemoth. This is one of the coolest things I’ve ever made with AI

I told Claude 4.6 Opus to make a pokemon clone - max effort It reasoned for 1 hour and 30 minutes and used 110k tokens and 2 shotted this absolute behemoth. This is one of the coolest things I’ve ever made with AI

Chris

658,874 views • 6 months ago

Claude 5 Fable (MAX) Created a MINECRAFT CLONE 1:1 Spent 55 Minutes/52.4k Tokens. It perfectly replicated the entire sandbox game: • Multiple biomes • Day/night cycle • Ores, caves & more This is actually insane. AI agents are getting scary good at game dev.

Claude 5 Fable (MAX) Created a MINECRAFT CLONE 1:1 Spent 55 Minutes/52.4k Tokens. It perfectly replicated the entire sandbox game: • Multiple biomes • Day/night cycle • Ores, caves & more This is actually insane. AI agents are getting scary good at game dev.

WorldofAI

350,390 views • 2 months ago

We are officially in the era where one prompt can ship a complete playable game in <50 bucks! Asked Claude Fable 5 and GPT 5.6 Sol to clone Terraria, max effort level Both are playable and offer a smooth experience GPT 5.6 Sol did it at roughly 1/4 the cost and in 2/3 the runtime: - GPT 5.6 Sol: ~$35, ~1h 20m - Claude Fable 5: ~$120, ~2h Tho Fable still produced the strongest results 🥲

We are officially in the era where one prompt can ship a complete playable game in <50 bucks! Asked Claude Fable 5 and GPT 5.6 Sol to clone Terraria, max effort level Both are playable and offer a smooth experience GPT 5.6 Sol did it at roughly 1/4 the cost and in 2/3 the runtime: - GPT 5.6 Sol: ~$35, ~1h 20m - Claude Fable 5: ~$120, ~2h Tho Fable still produced the strongest results 🥲

GMI Cloud

130,438 views • 1 month ago

Drake’s e-commerce store is insane 🤯 1. Instead of a traditional layout, his team built a visual map of his mansion with linked products in each room 2. He’s using Shopify new Collective feature to dropship custom collabs with select partner brands h/t Harley Finkelstein

Drake’s e-commerce store is insane 🤯 1. Instead of a traditional layout, his team built a visual map of his mansion with linked products in each room 2. He’s using Shopify new Collective feature to dropship custom collabs with select partner brands h/t Harley Finkelstein

Kallaway

205,094 views • 3 years ago

Everyone's using Seedance 2.0. Most people are prompting from scratch and re-generating til something works. We built promo templates so you can easily make professional videos: 1. Pick a promo video template 2. Paste a website URL 3. Done Every output is usable. No prompt engineering. No generating 30 times hoping one looks good. Try it now:

Everyone's using Seedance 2.0. Most people are prompting from scratch and re-generating til something works. We built promo templates so you can easily make professional videos: 1. Pick a promo video template 2. Paste a website URL 3. Done Every output is usable. No prompt engineering. No generating 30 times hoping one looks good. Try it now:

ChatCut

16,910 views • 5 months ago

.OpenAI's GPT-5.1 is live in Cline. After testing the last several weeks, here's what we've found makes GPT-5.1 an exceptional model in Cline: GPT-5.1 is an obsessive researcher. It reads everything before writing anything. Where other models grab context and go, GPT-5.1 investigates thoroughly first, even for one-line changes. We found this overkill for simple tasks, and modified GPT-5.1's prompting resultantly, but a superpower when channeled right. Here's what you should expect from GPT-5.1 in Cline: > deep, systematic investigation before any edits > follows constraints religiously (specs, todo lists, your requirements) > maintains coherent reasoning with large contexts > delivers complete features with fewer corrections How we optimized Cline for GPT-5.1: 1. redesigned prompts for execution focus and stricter Plan/Act transitions 2. tightened tool specs, GPT-5.1 internalizes rules with just single examples 3. split /deep-planning investigation into two phases (read first, then search) Result: GPT-5.1 is a model that finishes what it starts, even across dozens of files and thousands of lines. Both GPT-5.1 (272K context) and GPT-5.1-codex (400K context) available now in Cline. $1.25/M input, $10/M output. Enable native tool calling in settings for best results. Use /deep-planning to see GPT-5.1 at its best. Full technical breakdown linked below.

.OpenAI's GPT-5.1 is live in Cline. After testing the last several weeks, here's what we've found makes GPT-5.1 an exceptional model in Cline: GPT-5.1 is an obsessive researcher. It reads everything before writing anything. Where other models grab context and go, GPT-5.1 investigates thoroughly first, even for one-line changes. We found this overkill for simple tasks, and modified GPT-5.1's prompting resultantly, but a superpower when channeled right. Here's what you should expect from GPT-5.1 in Cline: > deep, systematic investigation before any edits > follows constraints religiously (specs, todo lists, your requirements) > maintains coherent reasoning with large contexts > delivers complete features with fewer corrections How we optimized Cline for GPT-5.1: 1. redesigned prompts for execution focus and stricter Plan/Act transitions 2. tightened tool specs, GPT-5.1 internalizes rules with just single examples 3. split /deep-planning investigation into two phases (read first, then search) Result: GPT-5.1 is a model that finishes what it starts, even across dozens of files and thousands of lines. Both GPT-5.1 (272K context) and GPT-5.1-codex (400K context) available now in Cline. $1.25/M input, $10/M output. Enable native tool calling in settings for best results. Use /deep-planning to see GPT-5.1 at its best. Full technical breakdown linked below.

Cline

72,111 views • 9 months ago

So I built a Google Lens mobile app using just the Shopify Catalog API (publicly available), GPT-5, Replit ⠕ and Claude R Perrin code. It’s insane how this was just built by telling it what I wanted. Here’s a demo. Should I open source? All APIs are publicly available. The app works like this: 1. Ask Replit to generate initial working app prototype. 2. Bring into Claude Code and iterate using screenshots and high level vision. The app works by: - user takes a photo or selects one from camera roll - crop photo to maximize focus area - send to GPT-5-mini to extract tags - send extracted product tags to gpt-4o-mini to generate search query - send search query to Shopify Catalog API Basically your camera becomes a shopping search bar.

So I built a Google Lens mobile app using just the Shopify Catalog API (publicly available), GPT-5, Replit ⠕ and Claude R Perrin code. It’s insane how this was just built by telling it what I wanted. Here’s a demo. Should I open source? All APIs are publicly available. The app works like this: 1. Ask Replit to generate initial working app prototype. 2. Bring into Claude Code and iterate using screenshots and high level vision. The app works by: - user takes a photo or selects one from camera roll - crop photo to maximize focus area - send to GPT-5-mini to extract tags - send extracted product tags to gpt-4o-mini to generate search query - send search query to Shopify Catalog API Basically your camera becomes a shopping search bar.

Gil

69,204 views • 7 months ago

built and deployed an AI Ad Analysis Tool on my vibe marketing resource site today... steps: 1) I've managed $150m+ in campaigns so used my knowledge to have Manus build out a super detailed scope using weights for various factors 2) loaded the scope into Replit and it basically one shotted the app! it uses reasoning models from OpenAI 3) went back and forth with Replit agent to update the overall design and UX 4) now we have tools, super cool. did some testing and everything is working with beehiiv Give it a try and lmk what you think... THIS IS VIBE MARKETING. Stack: - Replit - Manus - OpenAI - Beehive

built and deployed an AI Ad Analysis Tool on my vibe marketing resource site today... steps: 1) I've managed $150m+ in campaigns so used my knowledge to have Manus build out a super detailed scope using weights for various factors 2) loaded the scope into Replit and it basically one shotted the app! it uses reasoning models from OpenAI 3) went back and forth with Replit agent to update the overall design and UX 4) now we have tools, super cool. did some testing and everything is working with beehiiv Give it a try and lmk what you think... THIS IS VIBE MARKETING. Stack: - Replit - Manus - OpenAI - Beehive

The Boring Marketer

17,832 views • 1 year ago

Ruby doesn’t know she was surrendered. She doesn’t know a decision was made. She only knows her family disappeared. A volunteer took her out expecting to comfort a scared dog and ended up sitting on the floor with her while she pressed in close, buried her face, and quietly asked not to be left again. Ruby is a 1-year-old Shepherd / Belgian Malinois mix. Housebroken. Good with kids. Good with other dogs. She’s at Cobb County Animal Services in Marietta, GA. One share can be the difference between another night in a kennel and somebody saying come on girl, let’s go home. H/T Sharing.Lov Instagram

Ruby doesn’t know she was surrendered. She doesn’t know a decision was made. She only knows her family disappeared. A volunteer took her out expecting to comfort a scared dog and ended up sitting on the floor with her while she pressed in close, buried her face, and quietly asked not to be left again. Ruby is a 1-year-old Shepherd / Belgian Malinois mix. Housebroken. Good with kids. Good with other dogs. She’s at Cobb County Animal Services in Marietta, GA. One share can be the difference between another night in a kennel and somebody saying come on girl, let’s go home. H/T Sharing.Lov Instagram

German Shepherd Dad 🇺🇸 & 🔥

324,167 views • 3 months ago

🚨🚨INSANE FOOTAGE🚨🚨 Fake #49ers player Daejon Love allegedly sent women another video claiming he owned a $15M Malibu beach house, a $20M Beverly Hills home, a Miami beach house, and a Billionaire’s Row penthouse in Manhattan. He then showed off what he claimed were nearly $1 MILLION worth of car keys, including a Lamborghini, Corvette ZR1, Hellcat Redeye and Mercedes-AMG. Love also claimed he owned hundreds of thousands of dollars in jewelry, including a $300K Richard Mille watch. The alleged fake lifestyle just keeps getting crazier. (h/t Johnny)

🚨🚨INSANE FOOTAGE🚨🚨 Fake #49ers player Daejon Love allegedly sent women another video claiming he owned a $15M Malibu beach house, a $20M Beverly Hills home, a Miami beach house, and a Billionaire’s Row penthouse in Manhattan. He then showed off what he claimed were nearly $1 MILLION worth of car keys, including a Lamborghini, Corvette ZR1, Hellcat Redeye and Mercedes-AMG. Love also claimed he owned hundreds of thousands of dollars in jewelry, including a $300K Richard Mille watch. The alleged fake lifestyle just keeps getting crazier. (h/t Johnny)

MLFootball

1,340,173 views • 5 days ago

RFK Jr. Exposes Ukraine as a ‘Money Laundering Scheme’ for BlackRock This is insane. Robert F. Kennedy Jr reports: “Even if the Ukraine war ended today, we’re still going to spend half a trillion there rebuilding the country.” But get this, Mitch McConnell has essentially admitted that the money is not going to Ukraine, but to military contractors. That’s good news for General Dynamics, Boeing, and Lockheed Martin. But “who do you think owns every one of those companies?” Kennedy asked. The answer: BlackRock. “It’s all just a money laundering scheme.” H/T: Hambone

RFK Jr. Exposes Ukraine as a ‘Money Laundering Scheme’ for BlackRock This is insane. Robert F. Kennedy Jr reports: “Even if the Ukraine war ended today, we’re still going to spend half a trillion there rebuilding the country.” But get this, Mitch McConnell has essentially admitted that the money is not going to Ukraine, but to military contractors. That’s good news for General Dynamics, Boeing, and Lockheed Martin. But “who do you think owns every one of those companies?” Kennedy asked. The answer: BlackRock. “It’s all just a money laundering scheme.” H/T: Hambone

The Vigilant Fox 🦊

1,307,964 views • 2 years ago

Scaled Dot-Product Attention by hand ✍️ I explained attention in my recent AI seminar as follows: Step 1 — Compare We take one token as a query and compare it against all keys using dot products. This gives us a grid of similarity scores: every query against every key. Step 2 — Scale We divide those scores by √dₖ to keep the values numerically stable. Step 3 — Normalize We apply softmax to turn raw scores into a probability distribution—values between 0 and 1 that sum to 1. These are the attention weights. Step 4 — Combine We use those weights to compute a weighted sum of the values. Each output token becomes a linear combination of all previous tokens—maybe 37% from one, 23% from another, and a little from the rest. No blackbox. Just dot products, scaling, softmax, and weighted sums—built one row at a time in Excel.

Scaled Dot-Product Attention by hand ✍️ I explained attention in my recent AI seminar as follows: Step 1 — Compare We take one token as a query and compare it against all keys using dot products. This gives us a grid of similarity scores: every query against every key. Step 2 — Scale We divide those scores by √dₖ to keep the values numerically stable. Step 3 — Normalize We apply softmax to turn raw scores into a probability distribution—values between 0 and 1 that sum to 1. These are the attention weights. Step 4 — Combine We use those weights to compute a weighted sum of the values. Each output token becomes a linear combination of all previous tokens—maybe 37% from one, 23% from another, and a little from the rest. No blackbox. Just dot products, scaling, softmax, and weighted sums—built one row at a time in Excel.

Tom Yeh

31,114 views • 8 months ago

I told Claude to predict the next BTC candle using Kronos. didn't touch anything. +$4,200 overnight. Kronos isn't another LLM wrapper. it's a GPT-style model trained on 12 billion candles from 45 exchanges. Binance, NYSE, NASDAQ, LSE. it reads OHLC the way GPT reads English. > words = candle patterns > sentences = OHLC sequences > the corpus = every trade ever printed > the edge is simple: P(next | history) = softmax(W · h_t) same math GPT uses to predict the next word, except the "word" is the next candle. no twitter sentiment. no blog scraping. no hallucinated confidence. just statistics over 12B real price moves. I plugged Kronos into a 5-min BTC Polymarket bot. input: last 50 candles output: probability next candle closes up trades only when |p_model - p_market| > 0.04 72 hours: $300 → $4,500 ChatGPT looks at a chart and pattern-matches against an article it read. Kronos looks at the chart and remembers 12 billion of them. 87% of Polymarket is exit liquidity. one of these models is actually trading the market. 21.8k stars, free, MIT license. Tsinghua University release.

I told Claude to predict the next BTC candle using Kronos. didn't touch anything. +$4,200 overnight. Kronos isn't another LLM wrapper. it's a GPT-style model trained on 12 billion candles from 45 exchanges. Binance, NYSE, NASDAQ, LSE. it reads OHLC the way GPT reads English. > words = candle patterns > sentences = OHLC sequences > the corpus = every trade ever printed > the edge is simple: P(next | history) = softmax(W · h_t) same math GPT uses to predict the next word, except the "word" is the next candle. no twitter sentiment. no blog scraping. no hallucinated confidence. just statistics over 12B real price moves. I plugged Kronos into a 5-min BTC Polymarket bot. input: last 50 candles output: probability next candle closes up trades only when |p_model - p_market| > 0.04 72 hours: $300 → $4,500 ChatGPT looks at a chart and pattern-matches against an article it read. Kronos looks at the chart and remembers 12 billion of them. 87% of Polymarket is exit liquidity. one of these models is actually trading the market. 21.8k stars, free, MIT license. Tsinghua University release.

RetroChainer

94,059 views • 4 months ago

timelapse #86 (15 hrs): - got my first OOM on 8xB200 node - defaulting back to grok-code-fast-1, the fastest reliable coding model with by far most intuitive instruction following, combined with grok 4 fast reasoning to plan before i let grok code work its magic - drank 2 large tim hortons iced capps, loaded myself w/ creatine, daily nootropics - tried out gpt-5-codex but it simply doesnt match the speed i require when i go deep into one thing at a time sequentially - got caught watching youtube videos in the middle, need to make sure i block any and all content that could get in my way - caught up on all book revisions so getting super ahead with other chapters - developed an overnight addiction to switching color themes in cursor - did some pair programming w/ Kearm h/eng using Tuple on free trial - applying for O-1

timelapse #86 (15 hrs): - got my first OOM on 8xB200 node - defaulting back to grok-code-fast-1, the fastest reliable coding model with by far most intuitive instruction following, combined with grok 4 fast reasoning to plan before i let grok code work its magic - drank 2 large tim hortons iced capps, loaded myself w/ creatine, daily nootropics - tried out gpt-5-codex but it simply doesnt match the speed i require when i go deep into one thing at a time sequentially - got caught watching youtube videos in the middle, need to make sure i block any and all content that could get in my way - caught up on all book revisions so getting super ahead with other chapters - developed an overnight addiction to switching color themes in cursor - did some pair programming w/ Kearm h/eng using Tuple on free trial - applying for O-1

Elliot Arledge

103,006 views • 11 months ago

no coincidence here folks!!! There was a 1 in 625 trillion chance this photo of Lionel Messi bathing baby Lamine Yamal would ever happen. "Families entered a raffle, Barcelona players were assigned by random lottery, and somehow 20 year old Messi ended up pairing with this five month old baby, and that's Lamine Yamal right there." "The probability of this happening says you are more than 2 million times more likely to win the Powerball Jackpot than to have this script play out. It is roughly equivalent to picking one specific second out of 20 million years, 1 in 625 trillion, insane." H/T Laura Aboli

no coincidence here folks!!! There was a 1 in 625 trillion chance this photo of Lionel Messi bathing baby Lamine Yamal would ever happen. "Families entered a raffle, Barcelona players were assigned by random lottery, and somehow 20 year old Messi ended up pairing with this five month old baby, and that's Lamine Yamal right there." "The probability of this happening says you are more than 2 million times more likely to win the Powerball Jackpot than to have this script play out. It is roughly equivalent to picking one specific second out of 20 million years, 1 in 625 trillion, insane." H/T Laura Aboli

Paul White Gold Eagle

39,869 views • 1 month ago

$Auto regressive LLMs are officially on notice. run Gemma 4 26B diffusion gguf with llama.cpp Google just dropped DiffusionGemma-26B, and it completely flips how we generate text. instead of predicting words one by one, it generates 256 tokens in parallel using bi-directional attention. its like stable diffusion, but for language. the model starts with random text "noise" and iteratively refines and self-corrects the entire block in real-time to fix formatting and reasoning errors on the fly. since it’s a Mixture of Experts (MoE) that only activates 3.8B parameters during inference, it fits perfectly on consumer hardware. You can run the Q4_K_M quant with an 18GB VRAM budget on a single RTX 3090 or RTX 4090 with exceptional throughput. Tested on Ubuntu 22 with CUDA 13.1 using the cutting edge experimental llama.cpp branch. Here is how to compile and run it with the live terminal denoising visualizer: # 1. Clone & check out the experimental PR (#24423) - 1) git clone && cd llama.cpp -git fetch origin 2) pull/24423/head:diffusiongemma && --git checkout diffusiongemma # 2. Build with CUDA support 1) cmake -B build -DGGML_CUDA=ON -DCMAKE_CUDA_ARCHITECTURES=native 2) cmake --build build -j $(nproc) --config Release --target llama-diffusion-cli # 3. Run with live visual denoising (llama.cpp flags) ./build/bin/llama-diffusion-cli \ -m /path/to/diffusiongemma-26B-A4B-it-Q4_K_M.gguf \ -ngl 99 -cnv -n 2048 --diffusion-visual Watch the video below to see the live --diffusion-visual canvas iteratively de noising the prompt output in real time. guide and unsloth's hugging face GGUF model links are in the comments below! Is auto regressive generation officially legacy tech? Let me know what you think.$

Auto regressive LLMs are officially on notice. run Gemma 4 26B diffusion gguf with llama.cpp Google just dropped DiffusionGemma-26B, and it completely flips how we generate text. instead of predicting words one by one, it generates 256 tokens in parallel using bi-directional attention. its like stable diffusion, but for language. the model starts with random text "noise" and iteratively refines and self-corrects the entire block in real-time to fix formatting and reasoning errors on the fly. since it’s a Mixture of Experts (MoE) that only activates 3.8B parameters during inference, it fits perfectly on consumer hardware. You can run the Q4_K_M quant with an 18GB VRAM budget on a single RTX 3090 or RTX 4090 with exceptional throughput. Tested on Ubuntu 22 with CUDA 13.1 using the cutting edge experimental llama.cpp branch. Here is how to compile and run it with the live terminal denoising visualizer: # 1. Clone & check out the experimental PR (#24423) - 1) git clone && cd llama.cpp -git fetch origin 2) pull/24423/head:diffusiongemma && --git checkout diffusiongemma # 2. Build with CUDA support 1) cmake -B build -DGGML_CUDA=ON -DCMAKE_CUDA_ARCHITECTURES=native 2) cmake --build build -j $(nproc) --config Release --target llama-diffusion-cli # 3. Run with live visual denoising (llama.cpp flags) ./build/bin/llama-diffusion-cli \ -m /path/to/diffusiongemma-26B-A4B-it-Q4_K_M.gguf \ -ngl 99 -cnv -n 2048 --diffusion-visual Watch the video below to see the live --diffusion-visual canvas iteratively de noising the prompt output in real time. guide and unsloth's hugging face GGUF model links are in the comments below! Is auto regressive generation officially legacy tech? Let me know what you think.

Alok

52,656 views • 2 months ago