Loading video...

Video Failed to Load

Go Home

Tensor parallel loading in 🤗 Diffusers got a massive upgrade 🔥 On Flux.2-Dev DiT, with a TP degree of 4 (A10G): • 30.4s → 12.5s loading • 64.1 → 6.8 GB peak CPU memory/rank ~2.4× faster loading. ~89% less CPU memory. Check it out:

RisingSayak's profile picture

Sayak Paul

24,667 subscribers

13,129 views • 22 days ago •via X (Twitter)

LIVE

Anya Rossi• Live Now

Private livecam show

0 Comments

No comments available

Comments from the original post will appear here

Related Videos

How does Edge0 run a 35B model with only 1–2.5 GB of peak memory? In simple terms: • The full model stays in storage. • Only the parameters needed for the current step are loaded into memory. • The next parameters are prepared while the current step is running. • Memory is reused instead of loading the entire model at once. Large model. Small working set. Fully on-device.
0:10

How does Edge0 run a 35B model with only 1–2.5 GB of peak memory? In simple terms: • The full model stays in storage. • Only the parameters needed for the current step are loaded into memory. • The next parameters are prepared while the current step is running. • Memory is reused instead of loading the entire model at once. Large model. Small working set. Fully on-device.

Samuel Zeng

34,026 views • 1 month ago

You don't need a GPU for fast studio grade voice cloning anymore. Qwen3 TTS (1.7B Q4_K_M) + mainline llama.cpp is officially the fastest way to generate zero shot voice clones using 100% pure CPU execution. Following up on my last post where we ran the Q8 model on a GPU, we just took local C++ voice synthesis a massive step further. The open source community quantized Alibaba's SOTA Qwen3 TTS model down to Q4_K_M GGUF, completely freeing local audio pipelines from dedicated graphics hardware. Here is the real world benchmark and hardware breakdown of running SOTA voice cloning on CPU: # Architecture & Model Setup Using Qwen3-TTS-12Hz-1.7B-Base-Q4_K_M.gguf paired with the 8 bit multimodal projector (mmproj-Q8_0.gguf), llama.cpp executes the entire pipeline in pure C++. No PyTorch, no CUDA dependencies, and no VRAM bottlenecks. # Real-World Memory Footprint - Baseline RAM: 1.6 GB system idle. - Peak Generation RAM: 8 GB RAM during active voice synthesis. - Requirement: Any basic machine with at least 8 GB of system RAM can run this easily. # Real World CPU Benchmarks - Google Colab Free Tier (Throttled 2 Core CPU): Synthesizes a 5 sec studio quality audio clip (~8 words) in 45 seconds. - Modern Consumer CPU (Intel i5/i7 13th/14th Gen or AMD Ryzen 7000/9000): generation should drop to 5 to 20 seconds (nearly 1:1 real-time generation speed!). # Zero Shot Voice Cloning Quality Pass any 5 to 20 second .wav audio sample to the C++ engine using the --tts-speaker-file flag. It yields clean, natural sounding cloned speech with virtually zero quality loss compared to unquantized FP16 weights. To make testing seamless, I built an updated zero config Google Colab notebook. It pulls the official pre built llama.cpp CPU binaries (zero compilation time!) launches a live Gradio web app right in your browser. Record a 5 second clip from your mic (or drop a .mp3, .wav file), type text, and generate cloned audio on CPU. Native C++ audio models are making edge based, offline AI voice agents a reality. Links to the free Q4 CPU Colab notebook and the Q4_K_M GGUF HuggingFace repository are in the replies below! Which models have you been running on your CPUs? What CPU hardware are you using for local inference?
0:16

You don't need a GPU for fast studio grade voice cloning anymore. Qwen3 TTS (1.7B Q4_K_M) + mainline llama.cpp is officially the fastest way to generate zero shot voice clones using 100% pure CPU execution. Following up on my last post where we ran the Q8 model on a GPU, we just took local C++ voice synthesis a massive step further. The open source community quantized Alibaba's SOTA Qwen3 TTS model down to Q4_K_M GGUF, completely freeing local audio pipelines from dedicated graphics hardware. Here is the real world benchmark and hardware breakdown of running SOTA voice cloning on CPU: # Architecture & Model Setup Using Qwen3-TTS-12Hz-1.7B-Base-Q4_K_M.gguf paired with the 8 bit multimodal projector (mmproj-Q8_0.gguf), llama.cpp executes the entire pipeline in pure C++. No PyTorch, no CUDA dependencies, and no VRAM bottlenecks. # Real-World Memory Footprint - Baseline RAM: 1.6 GB system idle. - Peak Generation RAM: 8 GB RAM during active voice synthesis. - Requirement: Any basic machine with at least 8 GB of system RAM can run this easily. # Real World CPU Benchmarks - Google Colab Free Tier (Throttled 2 Core CPU): Synthesizes a 5 sec studio quality audio clip (~8 words) in 45 seconds. - Modern Consumer CPU (Intel i5/i7 13th/14th Gen or AMD Ryzen 7000/9000): generation should drop to 5 to 20 seconds (nearly 1:1 real-time generation speed!). # Zero Shot Voice Cloning Quality Pass any 5 to 20 second .wav audio sample to the C++ engine using the --tts-speaker-file flag. It yields clean, natural sounding cloned speech with virtually zero quality loss compared to unquantized FP16 weights. To make testing seamless, I built an updated zero config Google Colab notebook. It pulls the official pre built llama.cpp CPU binaries (zero compilation time!) launches a live Gradio web app right in your browser. Record a 5 second clip from your mic (or drop a .mp3, .wav file), type text, and generate cloned audio on CPU. Native C++ audio models are making edge based, offline AI voice agents a reality. Links to the free Q4 CPU Colab notebook and the Q4_K_M GGUF HuggingFace repository are in the replies below! Which models have you been running on your CPUs? What CPU hardware are you using for local inference?

Alok

105,756 views • 2 months ago

🌍 Back from hiatus! Continuing the Three.js Gamedev course with Infinite worlds—how do you keep them smooth? There's a bunch of moving parts here. You've got to keep an eye on memory, while loading and unloading assets smoothly and keeping framerate hitches to a minimum. In this section, we'll tackle how to build a streaming system and look at what features Three.js offers to make this easier. 🔔 Want to be notified? Sign up here 👉 #threejs #gamedev
0:10

🌍 Back from hiatus! Continuing the Three.js Gamedev course with Infinite worlds—how do you keep them smooth? There's a bunch of moving parts here. You've got to keep an eye on memory, while loading and unloading assets smoothly and keeping framerate hitches to a minimum. In this section, we'll tackle how to build a streaming system and look at what features Three.js offers to make this easier. 🔔 Want to be notified? Sign up here 👉 #threejs #gamedev

Simon

17,383 views • 1 year ago

🚨 JUST IN: Sen. Tommy Tuberville, next Governor of Alabama, NAILS IT after DHS confirms rents are dropping the most in states and cities who cooperate with ICE deportations "Turns out, deporting thousands of free-loading illegals is a GREAT thing for Americans. Who would’ve thought…DEPORT, DEPORT, DEPORT." 🔥 LFG! The state include: Arizona, Texas, Louisiana, Tennessee, Georgia and Florida COMMON THEME there. All gas no brakes on deportations, keep lowering the rent!
0:10

🚨 JUST IN: Sen. Tommy Tuberville, next Governor of Alabama, NAILS IT after DHS confirms rents are dropping the most in states and cities who cooperate with ICE deportations "Turns out, deporting thousands of free-loading illegals is a GREAT thing for Americans. Who would’ve thought…DEPORT, DEPORT, DEPORT." 🔥 LFG! The state include: Arizona, Texas, Louisiana, Tennessee, Georgia and Florida COMMON THEME there. All gas no brakes on deportations, keep lowering the rent!

Eric Daugherty

40,212 views • 24 days ago

MARCUS CHEN STACKED 30 MAC MINIS INTO AN AI SERVER FARM. ONE $599 MAC MINI REPLACES YOUR $200/MONTH CLAUDE CODE BILL WITH $3 IN ELECTRICITY two months ago a developer posted his claude code bill on reddit. $170 in 10 days. someone replied "i bought a mac mini m4. haven't paid anthropic since." apple stores ran out of mac minis the same week the m4 chip has 120 gb/s memory bandwidth and unified memory architecture. cpu and gpu share one pool so the model loads once and both read from it. a $599 mac mini runs ai faster than a $1,500 windows pc with a discrete gpu since january 2026 ollama supports the anthropic messages api format. claude code connects directly to your local mac mini with one environment variable. same interface, zero api costs, $0 per request a heavy developer pays $459 a month across claude code max, chatgpt pro, gemini, cursor and copilot. that's $5,508 a year. the mac mini pays off in 3 months and runs on $3 in electricity after that uber rolled out claude code to 5,000 engineers and burned through their $3.4 billion 2026 ai budget in 4 months. the people who own the hardware in 2026 are going to look very far ahead in 2028 bookmark this and read the article below
0:16

MARCUS CHEN STACKED 30 MAC MINIS INTO AN AI SERVER FARM. ONE $599 MAC MINI REPLACES YOUR $200/MONTH CLAUDE CODE BILL WITH $3 IN ELECTRICITY two months ago a developer posted his claude code bill on reddit. $170 in 10 days. someone replied "i bought a mac mini m4. haven't paid anthropic since." apple stores ran out of mac minis the same week the m4 chip has 120 gb/s memory bandwidth and unified memory architecture. cpu and gpu share one pool so the model loads once and both read from it. a $599 mac mini runs ai faster than a $1,500 windows pc with a discrete gpu since january 2026 ollama supports the anthropic messages api format. claude code connects directly to your local mac mini with one environment variable. same interface, zero api costs, $0 per request a heavy developer pays $459 a month across claude code max, chatgpt pro, gemini, cursor and copilot. that's $5,508 a year. the mac mini pays off in 3 months and runs on $3 in electricity after that uber rolled out claude code to 5,000 engineers and burned through their $3.4 billion 2026 ai budget in 4 months. the people who own the hardware in 2026 are going to look very far ahead in 2028 bookmark this and read the article below

starmex

358,026 views • 4 months ago

Run Gemma 4 26B MoE on 8GB VRAM with 250k context at 20+ tokens/sec If you own any 8GB VRAM graphics card, stop what you are doing. Local AI just had its absolute "Holy Shit" moment for budget hardware. Yesterday, I benchmarked Unsloth Gemma 4 12B Q4_K_XL on an 8GB card. The community went wild but immediately demanded more: "Can we run a 25B+ model on budget GPUs?" Today, I’m delivering exactly that. I am running a massive 26B parameter Mixture of Experts (MoE) model locally on a standard 8GB VRAM setup with 250k full native context!. If you own an RTX 3060, 3070, 4060, or any budget GPU with 8GB of VRAM, the local AI paradigm has completely changed. The performance metrics are astonishing: - 20 tokens/sec flat decode throughput. - Stable, flat decode speed even with massive prompts. - I threw a 60k token prompt at it, and it still clocked in at 20 TPS without dropping a single frame. # What about prefill? Yes, Time To First Token (TTFT) is slightly high when swallowing massive contexts. But with a solid 200 tokens/sec prefill speed, the wait is barely noticeable and highly usable. And this is running completely without Multi Token Prediction (MTP) active. How is this possible? It’s the magic of Google's new QAT (Quantization Aware Training) quants for Gemma 4. The model weight file (unsloth gemma-4-26B-A4B-it-qat-UD-Q4_K_XL.gguf) is only 13.2 GB, making it the ultimate local powerhouse. # The Test Setup: CPU: Intel Core i7 RAM: 16GB System RAM GPU: NVIDIA GeForce RTX 4060 Laptop GPU (8GB VRAM) # The Secret Sauce (The -cmoe Flag) To make this work properly on any 8GB card, you must use the -cmoe (CPU MoE) flag in llama.cpp. This flag isolates the heavy MoE expert weights directly to system memory (CPU/RAM) while letting your GPU focus strictly on the Attention layers and the KV Cache. It prevents VRAM spillage and holds the throughput rock solid. # The flags: -m "gemma-4-26B-A4B-it-qat-UD-Q4_K_XL.gguf" -cmoe -c 248000 -v Once running, just open the UI on localhost and toggle the new reasoning lightbulb icon in the text input box to watch the model perform multi step thinking. Are you still running smaller models, or are you ready to scale up your budget local setups? Let's discuss in the replies
0:19

Run Gemma 4 26B MoE on 8GB VRAM with 250k context at 20+ tokens/sec If you own any 8GB VRAM graphics card, stop what you are doing. Local AI just had its absolute "Holy Shit" moment for budget hardware. Yesterday, I benchmarked Unsloth Gemma 4 12B Q4_K_XL on an 8GB card. The community went wild but immediately demanded more: "Can we run a 25B+ model on budget GPUs?" Today, I’m delivering exactly that. I am running a massive 26B parameter Mixture of Experts (MoE) model locally on a standard 8GB VRAM setup with 250k full native context!. If you own an RTX 3060, 3070, 4060, or any budget GPU with 8GB of VRAM, the local AI paradigm has completely changed. The performance metrics are astonishing: - 20 tokens/sec flat decode throughput. - Stable, flat decode speed even with massive prompts. - I threw a 60k token prompt at it, and it still clocked in at 20 TPS without dropping a single frame. # What about prefill? Yes, Time To First Token (TTFT) is slightly high when swallowing massive contexts. But with a solid 200 tokens/sec prefill speed, the wait is barely noticeable and highly usable. And this is running completely without Multi Token Prediction (MTP) active. How is this possible? It’s the magic of Google's new QAT (Quantization Aware Training) quants for Gemma 4. The model weight file (unsloth gemma-4-26B-A4B-it-qat-UD-Q4_K_XL.gguf) is only 13.2 GB, making it the ultimate local powerhouse. # The Test Setup: CPU: Intel Core i7 RAM: 16GB System RAM GPU: NVIDIA GeForce RTX 4060 Laptop GPU (8GB VRAM) # The Secret Sauce (The -cmoe Flag) To make this work properly on any 8GB card, you must use the -cmoe (CPU MoE) flag in llama.cpp. This flag isolates the heavy MoE expert weights directly to system memory (CPU/RAM) while letting your GPU focus strictly on the Attention layers and the KV Cache. It prevents VRAM spillage and holds the throughput rock solid. # The flags: -m "gemma-4-26B-A4B-it-qat-UD-Q4_K_XL.gguf" -cmoe -c 248000 -v Once running, just open the UI on localhost and toggle the new reasoning lightbulb icon in the text input box to watch the model perform multi step thinking. Are you still running smaller models, or are you ready to scale up your budget local setups? Let's discuss in the replies

Alok

295,293 views • 4 months ago

1-bit Qwen3.8-27B just more than DOUBLED its MMLU-Pro score. They didn’t change the decoder. 29.04% -> 61.54%, according to a new paper from ISTA-DASLab. Same lab behind GSQ-RCO, some of the best tiny Qwen3.8-27B quants in my book. Their method is called Disaggregated Quantization. You use different weights to read the prompt and generate the answer. A trained NVFP4 prefiller processes your input and builds the cache. Then the tiny GGUF decoder takes over. They train the prefiller specifically for that decoder, so it learns to produce representations the heavily compressed model can use. The extra checkpoint is 12.8 GiB, but you don’t need all of it in VRAM. Stream it from SSD layer by layer, reusing GPU memory. With a long enough prompt, loading can overlap computation. That’s a pretty fucking good reason to pay attention if your GPU is short on memory. I want to test this on my 4x3090 rig with NVMe. The released implementation targets Blackwell, so this needs adaptation. First I’ll check whether the quality gains survive on Ampere, then measure whether offloading actually helps. Paper:
0:17

1-bit Qwen3.8-27B just more than DOUBLED its MMLU-Pro score. They didn’t change the decoder. 29.04% -> 61.54%, according to a new paper from ISTA-DASLab. Same lab behind GSQ-RCO, some of the best tiny Qwen3.8-27B quants in my book. Their method is called Disaggregated Quantization. You use different weights to read the prompt and generate the answer. A trained NVFP4 prefiller processes your input and builds the cache. Then the tiny GGUF decoder takes over. They train the prefiller specifically for that decoder, so it learns to produce representations the heavily compressed model can use. The extra checkpoint is 12.8 GiB, but you don’t need all of it in VRAM. Stream it from SSD layer by layer, reusing GPU memory. With a long enough prompt, loading can overlap computation. That’s a pretty fucking good reason to pay attention if your GPU is short on memory. I want to test this on my 4x3090 rig with NVMe. The released implementation targets Blackwell, so this needs adaptation. First I’ll check whether the quality gains survive on Ampere, then measure whether offloading actually helps. Paper:

Alexey Fateev

19,870 views • 24 days ago

This week’s ChatGPT feature drop - Aug 21: Another Friday, another roundup of what we shipped this week: 1/ Recent photos: Long press on + menu to quickly attach recents on iOS - this is a fun, power user feature. just beautiful design! also a handy shortcut. 2/ Time: We made it so ChatGPT better understands your current time. It was surprisingly bad at this before, for some interesting, complex reasons. Now it's better! 3/ Long convo loading: Long conversations load and run faster on Less waiting = better. 4/ 'No internet' errors: We show clearer updates in the UI when you’re waiting for an internet connection. It's so frustrating when ChatGPT times out. Now at least you'll know why, when it's the internet. thanks to the crew that keeps shipping every week and hope you all enjoy 🚀
0:11

This week’s ChatGPT feature drop - Aug 21: Another Friday, another roundup of what we shipped this week: 1/ Recent photos: Long press on + menu to quickly attach recents on iOS - this is a fun, power user feature. just beautiful design! also a handy shortcut. 2/ Time: We made it so ChatGPT better understands your current time. It was surprisingly bad at this before, for some interesting, complex reasons. Now it's better! 3/ Long convo loading: Long conversations load and run faster on Less waiting = better. 4/ 'No internet' errors: We show clearer updates in the UI when you’re waiting for an internet connection. It's so frustrating when ChatGPT times out. Now at least you'll know why, when it's the internet. thanks to the crew that keeps shipping every week and hope you all enjoy 🚀

Adam Fry

372,145 views • 2 months ago

Imagine having a ping pong robot! 🏓 Researchers and developers building physical AI: meet Reachy 2 from Pollen Robotics, an open-source, humanoid robot for real-world experimentation. It’s a bimanual mobile manipulator: each 7-DOF arm mimics human proportions and can lift up to 3 kg, giving dexterity for object handling. It can be controlled with Python and ROS2 Humble, or go straight into VR teleoperation, use a headset to move Reachy’s arms, hands, and head, and see through its cameras as if you’re in the robot’s own body. Want it to move around? A mobile base with three omnidirectional wheels, rich sensors, and LiDAR lets Reachy 2 navigate and explore its surroundings smoothly. 🗺️ Under the hood, it’s powered by a CPU system that’s ready for machine learning, perfect for loading AI frameworks and testing new models from Hugging Face directly on the robot. Keep making robots more, and more accessible Pollen team! ... and keep making more open source models to make robots more mainstream clem 🤗!
0:18

Imagine having a ping pong robot! 🏓 Researchers and developers building physical AI: meet Reachy 2 from Pollen Robotics, an open-source, humanoid robot for real-world experimentation. It’s a bimanual mobile manipulator: each 7-DOF arm mimics human proportions and can lift up to 3 kg, giving dexterity for object handling. It can be controlled with Python and ROS2 Humble, or go straight into VR teleoperation, use a headset to move Reachy’s arms, hands, and head, and see through its cameras as if you’re in the robot’s own body. Want it to move around? A mobile base with three omnidirectional wheels, rich sensors, and LiDAR lets Reachy 2 navigate and explore its surroundings smoothly. 🗺️ Under the hood, it’s powered by a CPU system that’s ready for machine learning, perfect for loading AI frameworks and testing new models from Hugging Face directly on the robot. Keep making robots more, and more accessible Pollen team! ... and keep making more open source models to make robots more mainstream clem 🤗!

Lukas Ziegler

37,221 views • 1 year ago

Holy shit... Microsoft open sourced an inference framework that runs a 100B parameter LLM on a single CPU. It's called BitNet. And it does what was supposed to be impossible. No GPU. No cloud. No $10K hardware setup. Just your laptop running a 100-billion parameter model at human reading speed. Here's how it works: Every other LLM stores weights in 32-bit or 16-bit floats. BitNet uses 1.58 bits. Weights are ternary just -1, 0, or +1. That's it. No floats. No expensive matrix math. Pure integer operations your CPU was already built for. The result: - 100B model runs on a single CPU at 5-7 tokens/second - 2.37x to 6.17x faster than llama.cpp on x86 - 82% lower energy consumption on x86 CPUs - 1.37x to 5.07x speedup on ARM (your MacBook) - Memory drops by 16-32x vs full-precision models The wildest part: Accuracy barely moves. BitNet b1.58 2B4T their flagship model was trained on 4 trillion tokens and benchmarks competitively against full-precision models of the same size. The quantization isn't destroying quality. It's just removing the bloat. What this actually means: - Run AI completely offline. Your data never leaves your machine - Deploy LLMs on phones, IoT devices, edge hardware - No more cloud API bills for inference - AI in regions with no reliable internet The model supports ARM and x86. Works on your MacBook, your Linux box, your Windows machine. 27.4K GitHub stars. 2.2K forks. Built by Microsoft Research. 100% Open Source. MIT License.
0:13

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

Guri Singh

2,180,357 views • 7 months ago

Indian Railways introduces a global game-changer for bulk cargo: Stainless Steel Containers. 🇮🇳 This isn't just a new box; it’s a logistics revolution. Check the numbers: 🔹 Unmatched Speed: Loaded 34.2 tonnes in just 15 mins | Unloaded in 5 mins. 🔹 Massive Payload: 68.4 tonnes per wagon (2 containers) --- matching BOXN capacity with better tech. 🔹 Full Rake Power: 48 wagons can move 3,300 tonnes of salt in a single trip. 🔹 Corrosion-Free: 20ft Stainless Steel body ensures zero rust and longer container life. 🔹 Zero Waste: Dual 7ft x 4ft top openings for precision mechanized loading. The Verdict: No manual cleaning, no tarpaulins, and lightning-fast turnaround times. This is the future of "Freight on Rail." #IndianRailways #Logistics #Innovation #MakeInIndia #WesternRailway #FreightEfficiency Ashwini Vaishnaw Ministry of Railways
0:13

Indian Railways introduces a global game-changer for bulk cargo: Stainless Steel Containers. 🇮🇳 This isn't just a new box; it’s a logistics revolution. Check the numbers: 🔹 Unmatched Speed: Loaded 34.2 tonnes in just 15 mins | Unloaded in 5 mins. 🔹 Massive Payload: 68.4 tonnes per wagon (2 containers) --- matching BOXN capacity with better tech. 🔹 Full Rake Power: 48 wagons can move 3,300 tonnes of salt in a single trip. 🔹 Corrosion-Free: 20ft Stainless Steel body ensures zero rust and longer container life. 🔹 Zero Waste: Dual 7ft x 4ft top openings for precision mechanized loading. The Verdict: No manual cleaning, no tarpaulins, and lightning-fast turnaround times. This is the future of "Freight on Rail." #IndianRailways #Logistics #Innovation #MakeInIndia #WesternRailway #FreightEfficiency Ashwini Vaishnaw Ministry of Railways

Ansuman Satapathy

19,977 views • 8 months ago

Researchers made KMeans 200x faster. And the new technique also beats approaches like cuML and FAISS. Flash-KMeans is an IO-aware implementation of exact KMeans that redesigns the algorithm around modern GPU bottlenecks. By attacking the memory bottlenecks directly, Flash-KMeans achieves: - 33x speedup over cuML - 200x speedup over FAISS This speedup comes from how it moves through GPU memory. Standard KMeans runs in two steps, and both are bottlenecked by reads and writes to GPU memory: 1) The first step matches every point to its nearest centroid. Standard KMeans computes the full point-to-centroid distance matrix, writes it out to GPU memory, then reads it back to find each nearest centroid. That write-then-read round trip is the bottleneck. Flash-KMeans combines the distance calculation with the nearest-centroid step, so the result is computed on-chip and the full matrix is never written out. 2) The second step recomputes each centroid by averaging the points assigned to it. Standard KMeans has thousands of threads writing into the same centroid slots at once, so they stall waiting for their turn. Flash-KMeans sorts points by cluster first, turning scattered writes into sequential reductions that read and write memory in one efficient pass. Using these two optimizations at the million-scale, Flash-KMeans completes a standard KMeans iteration in a few milliseconds. The video below depicts this in action. Several reasons why this is important: KMeans has always been an offline primitive. Something you run once to preprocess data and move on. These speedups make the approach viable in several runtime-critical systems. ↳ Vector indices like FAISS use KMeans to build search indices. Faster KMeans means you can re-index dynamically as data changes. ↳ LLM quantization methods need KMeans to find optimal weight codebooks, per layer, repeatedly. What takes hours could now take minutes. ↳ MoE models need fast token routing at inference time. Flash-KMeans makes it viable to run this inside the inference loop, not just in preprocessing. I have shared the paper in the replies. That said, memory is the real constraint Flash-KMeans solves, and the problem is not just limited to clustering. The vectors a RAG system stores after indexing create similar bottlenecks. I wrote a detailed walkthrough recently on cutting this vector memory by 32x with binary quantization, querying 36M+ vectors in a few milliseconds. Read it below.
0:12

Researchers made KMeans 200x faster. And the new technique also beats approaches like cuML and FAISS. Flash-KMeans is an IO-aware implementation of exact KMeans that redesigns the algorithm around modern GPU bottlenecks. By attacking the memory bottlenecks directly, Flash-KMeans achieves: - 33x speedup over cuML - 200x speedup over FAISS This speedup comes from how it moves through GPU memory. Standard KMeans runs in two steps, and both are bottlenecked by reads and writes to GPU memory: 1) The first step matches every point to its nearest centroid. Standard KMeans computes the full point-to-centroid distance matrix, writes it out to GPU memory, then reads it back to find each nearest centroid. That write-then-read round trip is the bottleneck. Flash-KMeans combines the distance calculation with the nearest-centroid step, so the result is computed on-chip and the full matrix is never written out. 2) The second step recomputes each centroid by averaging the points assigned to it. Standard KMeans has thousands of threads writing into the same centroid slots at once, so they stall waiting for their turn. Flash-KMeans sorts points by cluster first, turning scattered writes into sequential reductions that read and write memory in one efficient pass. Using these two optimizations at the million-scale, Flash-KMeans completes a standard KMeans iteration in a few milliseconds. The video below depicts this in action. Several reasons why this is important: KMeans has always been an offline primitive. Something you run once to preprocess data and move on. These speedups make the approach viable in several runtime-critical systems. ↳ Vector indices like FAISS use KMeans to build search indices. Faster KMeans means you can re-index dynamically as data changes. ↳ LLM quantization methods need KMeans to find optimal weight codebooks, per layer, repeatedly. What takes hours could now take minutes. ↳ MoE models need fast token routing at inference time. Flash-KMeans makes it viable to run this inside the inference loop, not just in preprocessing. I have shared the paper in the replies. That said, memory is the real constraint Flash-KMeans solves, and the problem is not just limited to clustering. The vectors a RAG system stores after indexing create similar bottlenecks. I wrote a detailed walkthrough recently on cutting this vector memory by 32x with binary quantization, querying 36M+ vectors in a few milliseconds. Read it below.

Avi Chawla

90,078 views • 4 months ago

This Chinese developer launched Llama 70B locally on a MacBook on a plane and for a full 11 hours without internet ran client projects. He was sitting by the window on a transatlantic flight with a MacBook Pro M4 with 64 GB of memory. WiFi on board cost $25 for the flight. He declined. No cloud API, no connection to Anthropic or OpenAI servers, no internet at all. Just a local Llama 3.3 70B on bf16 and his own orchestrator script. The model runs through llama.cpp. Generation speed, 71 tokens per second. Context around 60,000 tokens. Memory usage, 48.6 GiB out of 64. Battery at takeoff, 3 hours 21 minutes. And he gave the orchestrator this system prompt before takeoff: "You are an offline orchestrator running on a single MacBook. There is no network. The only resources you have are local files in /Users/dev/work, the Llama 70B inference server at localhost:8080, and a battery budget of 3 hours 21 minutes. Process the queue at /Users/dev/work/queue.jsonl (one client task per line). For each task: draft → run local evals → save artefact to /Users/dev/work/done/. Save context checkpoints every 12 tasks so you can resume after a battery swap. Stop only on empty queue or when battery drops below 5%." So the system knows exactly what resources it is running on. It knows it has no connection to the outside world for the next 11 hours. It knows it has finite memory and a finite battery. It knows the human will not intervene until the plane lands. The system runs in 1 loop. Takes a task from the queue, runs it through inference, saves the artifact, writes a checkpoint. Task after task, just like that. And only when the battery drops below 5% does the orchestrator automatically pause, waits for the laptop to switch to the backup power bank, and continues from the last checkpoint. Here is what the system actually writes in his log during the flight: "saved context checkpoint 8 of 12 (pos_min = 488, pos_max = 50118, size = 62.813 MiB)" "restored context checkpoint (pos_min = 488, pos_max = 50118)" "prompt processing progress: n_tokens = 50 / 60 818" "task 37016 done | tps = 71 s tokens text → /Users/dev/work/done/proposal_westside.md" Outside the window, clouds, blue sky, and no WiFi. On the tray, 1 MacBook, an open terminal on 2 screens, and an inference server on localhost. From what I have observed, this is the cleanest offline AI workflow I have seen in the past year: 11 hours of flight, $0 for WiFi, and the entire client queue closed before landing.
0:10

This Chinese developer launched Llama 70B locally on a MacBook on a plane and for a full 11 hours without internet ran client projects. He was sitting by the window on a transatlantic flight with a MacBook Pro M4 with 64 GB of memory. WiFi on board cost $25 for the flight. He declined. No cloud API, no connection to Anthropic or OpenAI servers, no internet at all. Just a local Llama 3.3 70B on bf16 and his own orchestrator script. The model runs through llama.cpp. Generation speed, 71 tokens per second. Context around 60,000 tokens. Memory usage, 48.6 GiB out of 64. Battery at takeoff, 3 hours 21 minutes. And he gave the orchestrator this system prompt before takeoff: "You are an offline orchestrator running on a single MacBook. There is no network. The only resources you have are local files in /Users/dev/work, the Llama 70B inference server at localhost:8080, and a battery budget of 3 hours 21 minutes. Process the queue at /Users/dev/work/queue.jsonl (one client task per line). For each task: draft → run local evals → save artefact to /Users/dev/work/done/. Save context checkpoints every 12 tasks so you can resume after a battery swap. Stop only on empty queue or when battery drops below 5%." So the system knows exactly what resources it is running on. It knows it has no connection to the outside world for the next 11 hours. It knows it has finite memory and a finite battery. It knows the human will not intervene until the plane lands. The system runs in 1 loop. Takes a task from the queue, runs it through inference, saves the artifact, writes a checkpoint. Task after task, just like that. And only when the battery drops below 5% does the orchestrator automatically pause, waits for the laptop to switch to the backup power bank, and continues from the last checkpoint. Here is what the system actually writes in his log during the flight: "saved context checkpoint 8 of 12 (pos_min = 488, pos_max = 50118, size = 62.813 MiB)" "restored context checkpoint (pos_min = 488, pos_max = 50118)" "prompt processing progress: n_tokens = 50 / 60 818" "task 37016 done | tps = 71 s tokens text → /Users/dev/work/done/proposal_westside.md" Outside the window, clouds, blue sky, and no WiFi. On the tray, 1 MacBook, an open terminal on 2 screens, and an inference server on localhost. From what I have observed, this is the cleanest offline AI workflow I have seen in the past year: 11 hours of flight, $0 for WiFi, and the entire client queue closed before landing.

Blaze

1,843,280 views • 5 months ago

🚨 Anthropic committed up to 1M TPU chips for Claude. Openai is leasing TPUs for chatgpt inference. Here's How kernels work on TPUs (deep dive 2/6 by emilio andere) pallas is Google's answer to kernel writing. a python kernel SDK built on JAX. still very experimental (jax.experimental.pallas). on TPU it compiles through mosaic; on GPU it lowers to triton. if you know CUDA, the syntax will feel familiar but the execution model is completely different. in CUDA, grid=(4,4) launches 16 blocks running simultaneously across SMs. in pallas, those 16 iterations run one after another in lexicographic order. no threads. no warps. no blocks. no occupancy tuning. a TPU is a sequential machine with a very wide vector register — more like a CPU than a GPU. performance comes from width: a 128x128 systolic array doing matmul and an 8x128 SIMD vector unit doing everything else. maximum parallelism on chip: 2, one per TensorCore in megacore mode. three concepts replace CUDA's thread/block/grid hierarchy. Refs are mutable memory references. because execution is sequential, each iteration safely accumulates without atomics. in CUDA you'd need atomics or a separate reduction pass. the memory model is also very different from NVIDIA's. zero hardware caches. VMEM is 32-128 MiB of software-managed scratchpad — 500-1000x larger than GPU shared memory per SM. all data must be explicitly DMA'd from HBM to VMEM before any computation touches it. four levels: HBM → VMEM → VREGs → MXU/VPU, plus SMEM for scalar control data. every byte of data movement is your responsibility. this is like CUDA shared memory except it's 500x bigger and there's no cache fallback. pipelining is mandatory. without double-buffering HBM→VMEM transfers, the MXU just stalls waiting for data. this is the single most important optimization on TPU. and because grid execution is sequential and deterministic, consecutive iterations that need the same input block skip the redundant HBM transfer automatically, impossible on GPU where block execution order is undefined. the compilation pipeline is unlike anything in this series: python → jaxpr → stableHLO → XLA HLO (71+ optimization passes) → LLO (78+ passes) → 322-bit VLIW bundles. the compiler packs instructions for scalar, vector, matrix, and DMA units into a single 322-bit word. everything in that bundle executes in parallel, with no runtime scheduling.
0:18

🚨 Anthropic committed up to 1M TPU chips for Claude. Openai is leasing TPUs for chatgpt inference. Here's How kernels work on TPUs (deep dive 2/6 by emilio andere) pallas is Google's answer to kernel writing. a python kernel SDK built on JAX. still very experimental (jax.experimental.pallas). on TPU it compiles through mosaic; on GPU it lowers to triton. if you know CUDA, the syntax will feel familiar but the execution model is completely different. in CUDA, grid=(4,4) launches 16 blocks running simultaneously across SMs. in pallas, those 16 iterations run one after another in lexicographic order. no threads. no warps. no blocks. no occupancy tuning. a TPU is a sequential machine with a very wide vector register — more like a CPU than a GPU. performance comes from width: a 128x128 systolic array doing matmul and an 8x128 SIMD vector unit doing everything else. maximum parallelism on chip: 2, one per TensorCore in megacore mode. three concepts replace CUDA's thread/block/grid hierarchy. Refs are mutable memory references. because execution is sequential, each iteration safely accumulates without atomics. in CUDA you'd need atomics or a separate reduction pass. the memory model is also very different from NVIDIA's. zero hardware caches. VMEM is 32-128 MiB of software-managed scratchpad — 500-1000x larger than GPU shared memory per SM. all data must be explicitly DMA'd from HBM to VMEM before any computation touches it. four levels: HBM → VMEM → VREGs → MXU/VPU, plus SMEM for scalar control data. every byte of data movement is your responsibility. this is like CUDA shared memory except it's 500x bigger and there's no cache fallback. pipelining is mandatory. without double-buffering HBM→VMEM transfers, the MXU just stalls waiting for data. this is the single most important optimization on TPU. and because grid execution is sequential and deterministic, consecutive iterations that need the same input block skip the redundant HBM transfer automatically, impossible on GPU where block execution order is undefined. the compilation pipeline is unlike anything in this series: python → jaxpr → stableHLO → XLA HLO (71+ optimization passes) → LLO (78+ passes) → 322-bit VLIW bundles. the compiler packs instructions for scalar, vector, matrix, and DMA units into a single 322-bit word. everything in that bundle executes in parallel, with no runtime scheduling.

wafer

33,655 views • 3 months ago

Another WTF moment. A developer just open-sourced a coding agent harness that boots 245x faster than Claude Code. It's called jcode. You launch it and the first frame renders in 14 milliseconds. Claude Code takes 3,436. One active session uses 27.8 MB of RAM. Claude Code uses 386.6. Run ten sessions in parallel and jcode holds at 117 MB while OpenCode swells to 3.2 GB. Each agent has a semantic memory graph instead of a scratchpad. Every turn gets embedded as a vector. The graph is queried on every turn for related memories, and a sideagent verifies the hits before injecting them into context. Consolidation runs in the background to check for stale or conflicting facts. No manual /remember calls. No token burn on lookup tools. The provider list is 30+ deep. Claude, ChatGPT, Gemini, GitHub Copilot, Azure, OpenRouter, DeepSeek, Groq, Mistral, Perplexity, Fireworks, Ollama, LM Studio, and any OpenAI-compatible endpoint you point it at. Ran out of tokens on your first ChatGPT Pro sub? /account swaps to the second. Then there's Swarm. Spawn two agents in the same repo and the server manages them. When agent A edits a file agent B has been reading, agent B gets pinged and can check the diff. Agents can DM each other, broadcast to the room, or spawn their own worker teams for parallel tasks. Groups, channels, and completion statuses are handled automatically. The UI has live side panels that render mermaid diagrams inline. To make it fast, the author wrote a Rust mermaid renderer 1800x faster than the JavaScript one, then wrote a custom terminal called Handterm because no existing terminal could do smooth partial-line scrolling. Self-dev mode is where it gets wild. Tell your agent to enter self-dev and it starts editing jcode's own source code, rebuilds the binary, reloads it live, and keeps working across your existing sessions. You can also resume broken sessions from Claude Code, Codex, OpenCode, or pi directly inside jcode. Anthropic's cache goes cold at the 5-minute mark and you're staring down a big cache miss on your next turn? The UI warns you before you spend the tokens. Written in Rust. MIT licensed. Runs on macOS, Windows, Linux, and Termux. Sitting at 11.2k stars with a native iOS app coming.
0:17

Another WTF moment. A developer just open-sourced a coding agent harness that boots 245x faster than Claude Code. It's called jcode. You launch it and the first frame renders in 14 milliseconds. Claude Code takes 3,436. One active session uses 27.8 MB of RAM. Claude Code uses 386.6. Run ten sessions in parallel and jcode holds at 117 MB while OpenCode swells to 3.2 GB. Each agent has a semantic memory graph instead of a scratchpad. Every turn gets embedded as a vector. The graph is queried on every turn for related memories, and a sideagent verifies the hits before injecting them into context. Consolidation runs in the background to check for stale or conflicting facts. No manual /remember calls. No token burn on lookup tools. The provider list is 30+ deep. Claude, ChatGPT, Gemini, GitHub Copilot, Azure, OpenRouter, DeepSeek, Groq, Mistral, Perplexity, Fireworks, Ollama, LM Studio, and any OpenAI-compatible endpoint you point it at. Ran out of tokens on your first ChatGPT Pro sub? /account swaps to the second. Then there's Swarm. Spawn two agents in the same repo and the server manages them. When agent A edits a file agent B has been reading, agent B gets pinged and can check the diff. Agents can DM each other, broadcast to the room, or spawn their own worker teams for parallel tasks. Groups, channels, and completion statuses are handled automatically. The UI has live side panels that render mermaid diagrams inline. To make it fast, the author wrote a Rust mermaid renderer 1800x faster than the JavaScript one, then wrote a custom terminal called Handterm because no existing terminal could do smooth partial-line scrolling. Self-dev mode is where it gets wild. Tell your agent to enter self-dev and it starts editing jcode's own source code, rebuilds the binary, reloads it live, and keeps working across your existing sessions. You can also resume broken sessions from Claude Code, Codex, OpenCode, or pi directly inside jcode. Anthropic's cache goes cold at the 5-minute mark and you're staring down a big cache miss on your next turn? The UI warns you before you spend the tokens. Written in Rust. MIT licensed. Runs on macOS, Windows, Linux, and Termux. Sitting at 11.2k stars with a native iOS app coming.

Brady Long

209,050 views • 2 months ago

Gemma 4 26B A4B MoE - 500+ t/s decode - Single RTX 4090 (24 GB VRAM) - Llama.cpp concurrency 24 - q8 kv cache How many API users can you simultaneously host on a single RTX 4090 (24 GB VRAM) before it crashes? Yesterday, I proved you can host 14 active users using unquantized memory. Today, I used 8 bit KV Cache Quantization to hack the VRAM footprint. I successfully scaled to 24 concurrent users without a single dropped connection. A 71% server capacity boost for free. By adding the -ctk q8_0 -ctv q8_0 flags to llama.cpp, you compress the KV cache context memory from 16 bit to 8 bit. This unlocks massive concurrency limits on Gemma 4 26B (MoE) on a single 24GB consumer GPU. Here is the exact telemetry from pushing 8 bit quantization to its absolute physical edge: # TEST 1: The 24 User Concurrency Max Server Config: 24 slots (np 24) | 4,096 context per slot | 98,304 Total Context Client Load: 24 simultaneous requests (2,000 token prompt per user) Unquantized KV cache for this load requires 28GB+ VRAM (Instant OOM). Quantized to Q8, it allocated safely at 23.35 GB. The C++ engine crunched the entire batch in 28.5 seconds. Decode Speed: 21 t/s (Per User) | 500 t/s (Agg) # TEST 2: The 48 User Queue Overload What happens to a compressed cache during a traffic spike? Server Config: 24 slots (np 24) | 4,096 context per slot | 98,304 Total Context Client Load: 48 simultaneous requests (2k token prompt per user) Zero queue drops. The scheduler flushed and hot swapped the 8 bit memory flawlessly on the fly, completing all 48 users in 66.0 seconds (a perfect 2.3x queue scaling multiplier). Decode Speed: 18 t/s (Per User) | 430 t/s (Agg) # TEST 3: The 8 User RAG Slam Server Config: 8 slots (np 8) | 60,000 context per slot | 480,000 Total Context Client Load: 8 simultaneous requests (30k token prompt per user) It allocated 23.83 GB VRAM and chewed through ~240,000 prefill tokens in 46 seconds under massive memory pressure. Prefill Speed: 6,200 t/s (Agg) Decode Speed: 22 t/s (Per User) | 175 t/s (Agg) # The Engineering Alpha (The Quantization Tradeoff): You gain a massive 71% increase in server capacity, but what do you lose? Compute latency. Because the cache is stored in 8 bit, the GPU's cores have to dequantize the memory back to 16 bit on the fly during every single prefill step. In my unquantized tests yesterday, single slot prefill was hitting ~1,500+ t/s. Today, under the heavy 48-user Q8 load, prefill dropped as low as ~750 t/s. You trade a few seconds of initial prefill latency to essentially double your API hosting capacity. For production high volume SaaS, this is the ultimate unit economics cheat code. Here is the exact command to run a 24 user Q8 continuous batching server on your own single 4090, single 3090 or any 24gb vram rig: ./build/bin/llama-server -m gemma-4-26B-A4B-it.gguf -c 98304 -np 24 -b 2048 -ub 2048 -ngl 99 -fa on -ctk q8_0 -ctv q8_0 --port 8080 (Note: -c 98304 allocates exactly 4,096 tokens of context per user across 24 slots). Hugging Face links to the Unsloth Gemma 4 26B QAT quants along with performance graphs available in the replies. Would you trade 3 seconds of Time To First Token latency to double your active user capacity?
0:18

Gemma 4 26B A4B MoE - 500+ t/s decode - Single RTX 4090 (24 GB VRAM) - Llama.cpp concurrency 24 - q8 kv cache How many API users can you simultaneously host on a single RTX 4090 (24 GB VRAM) before it crashes? Yesterday, I proved you can host 14 active users using unquantized memory. Today, I used 8 bit KV Cache Quantization to hack the VRAM footprint. I successfully scaled to 24 concurrent users without a single dropped connection. A 71% server capacity boost for free. By adding the -ctk q8_0 -ctv q8_0 flags to llama.cpp, you compress the KV cache context memory from 16 bit to 8 bit. This unlocks massive concurrency limits on Gemma 4 26B (MoE) on a single 24GB consumer GPU. Here is the exact telemetry from pushing 8 bit quantization to its absolute physical edge: # TEST 1: The 24 User Concurrency Max Server Config: 24 slots (np 24) | 4,096 context per slot | 98,304 Total Context Client Load: 24 simultaneous requests (2,000 token prompt per user) Unquantized KV cache for this load requires 28GB+ VRAM (Instant OOM). Quantized to Q8, it allocated safely at 23.35 GB. The C++ engine crunched the entire batch in 28.5 seconds. Decode Speed: 21 t/s (Per User) | 500 t/s (Agg) # TEST 2: The 48 User Queue Overload What happens to a compressed cache during a traffic spike? Server Config: 24 slots (np 24) | 4,096 context per slot | 98,304 Total Context Client Load: 48 simultaneous requests (2k token prompt per user) Zero queue drops. The scheduler flushed and hot swapped the 8 bit memory flawlessly on the fly, completing all 48 users in 66.0 seconds (a perfect 2.3x queue scaling multiplier). Decode Speed: 18 t/s (Per User) | 430 t/s (Agg) # TEST 3: The 8 User RAG Slam Server Config: 8 slots (np 8) | 60,000 context per slot | 480,000 Total Context Client Load: 8 simultaneous requests (30k token prompt per user) It allocated 23.83 GB VRAM and chewed through ~240,000 prefill tokens in 46 seconds under massive memory pressure. Prefill Speed: 6,200 t/s (Agg) Decode Speed: 22 t/s (Per User) | 175 t/s (Agg) # The Engineering Alpha (The Quantization Tradeoff): You gain a massive 71% increase in server capacity, but what do you lose? Compute latency. Because the cache is stored in 8 bit, the GPU's cores have to dequantize the memory back to 16 bit on the fly during every single prefill step. In my unquantized tests yesterday, single slot prefill was hitting ~1,500+ t/s. Today, under the heavy 48-user Q8 load, prefill dropped as low as ~750 t/s. You trade a few seconds of initial prefill latency to essentially double your API hosting capacity. For production high volume SaaS, this is the ultimate unit economics cheat code. Here is the exact command to run a 24 user Q8 continuous batching server on your own single 4090, single 3090 or any 24gb vram rig: ./build/bin/llama-server -m gemma-4-26B-A4B-it.gguf -c 98304 -np 24 -b 2048 -ub 2048 -ngl 99 -fa on -ctk q8_0 -ctv q8_0 --port 8080 (Note: -c 98304 allocates exactly 4,096 tokens of context per user across 24 slots). Hugging Face links to the Unsloth Gemma 4 26B QAT quants along with performance graphs available in the replies. Would you trade 3 seconds of Time To First Token latency to double your active user capacity?

Alok

17,465 views • 2 months ago

🇹🇷Turkish media is in complete panic mode over a 🇬🇷Greek newspaper report. Today, a Greek newspaper published a 4-step plan to hit Turkey first. The paper says: if Turkey starts moving troops to the coast, loading landing ships, and putting missiles on high alert, Greece treats that as an attack that is about to start. Then Greece strikes first. First, they watch with satellites and allied intelligence. Next, they try to blind Turkish radars and cut communications. Then warplanes hit airbases like Dalaman, Balıkesir, and Çiğli. Missiles from Greek islands hit landing ships while they are still in port. After that, the US, NATO, and the EU push for a ceasefire- and Greece sits at the table from a stronger position. The paper says this kind of first strike could decide the fight in less than 24 hours. Turkish media is already calling it a shock attack plan.
0:16

🇹🇷Turkish media is in complete panic mode over a 🇬🇷Greek newspaper report. Today, a Greek newspaper published a 4-step plan to hit Turkey first. The paper says: if Turkey starts moving troops to the coast, loading landing ships, and putting missiles on high alert, Greece treats that as an attack that is about to start. Then Greece strikes first. First, they watch with satellites and allied intelligence. Next, they try to blind Turkish radars and cut communications. Then warplanes hit airbases like Dalaman, Balıkesir, and Çiğli. Missiles from Greek islands hit landing ships while they are still in port. After that, the US, NATO, and the EU push for a ceasefire- and Greece sits at the table from a stronger position. The paper says this kind of first strike could decide the fight in less than 24 hours. Turkish media is already calling it a shock attack plan.

World updates

331,996 views • 25 days ago

i left my Grok Bot running overnight and woke up to a warehouse i never built. one prompt. i typed "just start selling stuff for me" and closed the laptop. that was it. no store, no supplier, no product picked by morning there was a whole floor humming. aisles stocked - SCOUT bot digging winners - SOURCE bot linking suppliers - PICK and PACK working the benches - ADS buying traffic - CARE clearing returns - SHIP bot loading vans that just kept rolling out the door all of it feeding one glowing hub in the middle. orders routed, live, while i slept the part that actually got me: i never messaged a supplier. never printed a label. never approved a single order. it tested products on its own, destroyed the losers, restocked the winners overnight, and shipped end to end with my phone face down on the nightstand $15 of ad spend in. a running store out. LIVE ORDERS chart pinned green, +128% on the fund, vans out the door before i'd had coffee i didn't find a product to sell. i hired a floor and went to sleep. that's the part nobody's ready for the store doesn't wait for me anymore. it just runs wanna guide? bookmark this and follow me
0:17

i left my Grok Bot running overnight and woke up to a warehouse i never built. one prompt. i typed "just start selling stuff for me" and closed the laptop. that was it. no store, no supplier, no product picked by morning there was a whole floor humming. aisles stocked - SCOUT bot digging winners - SOURCE bot linking suppliers - PICK and PACK working the benches - ADS buying traffic - CARE clearing returns - SHIP bot loading vans that just kept rolling out the door all of it feeding one glowing hub in the middle. orders routed, live, while i slept the part that actually got me: i never messaged a supplier. never printed a label. never approved a single order. it tested products on its own, destroyed the losers, restocked the winners overnight, and shipped end to end with my phone face down on the nightstand $15 of ad spend in. a running store out. LIVE ORDERS chart pinned green, +128% on the fund, vans out the door before i'd had coffee i didn't find a product to sell. i hired a floor and went to sleep. that's the part nobody's ready for the store doesn't wait for me anymore. it just runs wanna guide? bookmark this and follow me

cristal💎

96,198 views • 1 month ago