Depth Anything 3 now runs as pure C++/ggml (ggml)... . No Python, no PyTorch, no CUDA toolkit at inference, just one self-contained GGUF. It's faster than PyTorch on CPU! and ties speed on GPU. The CPU win came from the last place..I'd have looked. Quantized GGUF on Hugging Face🤗 Shout out to Georgi Gerganov for ggml (we are building a ggml-world!❤️) and to ByteDance Open Source and Depth Anything 3 authors Bingyi Kang Jun Hao Liew Donny Y. Chen !show more

parakeet.cpp: native C++/ggml (ggml) inference for NVIDIA AI Developer's... Parakeet, one of the best speech-to-text models out there, from the LocalAI team. Every Parakeet model (TDT/CTC/RNNT/hybrid + cache-aware streaming), byte-for-byte identical output to NeMo, now running anywhere with no Python and even a bit faster, on CPU and GPU. Quantized GGUF on Hugging Face 🤗 Huge thanks to Georgi Gerganov for ggml and to NVIDIA AI Developer for releasing Parakeet! 🧵show more

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.show more

First fully ML-framework-free 3D Gaussian Splatting implementation in LichtFeld... Studio. I’ve completed the migration of the full training pipeline to a custom CUDA-based tensor library. No PyTorch, no LibTorch, no autograd. Every gradient is implemented by hand, either through CUDA kernels or minimal abstractions on top. This makes it the first full training setup for 3D Gaussian Splatting with zero dependencies on existing ML frameworks. It’s not just about independence, it's about control! We now manage every byte of GPU memory, which opens the door to tighter optimization and finer performance tuning. The framework footprint is minimal, without pulling in gigabytes of ML runtime code that was never designed for real-time or graphics-driven applications. A few modules, such as the metrics and 3DGUT interfaces, are still being ported, and some operations are temporarily naïve, so performance is not yet on par with master. But this refactor lays the groundwork for: - A fully self-contained binary - Fine-grained memory optimization - Easier experimentation without the weight of an ML stack We’re getting close.show more

90% of "AI developers" just download pre packaged GGUF... files from Hugging Face, hit run, and call it a day. The top 10% know how to pull the raw safetensors, run the math, and quantize massive models into Q4_K_M themselves. If you think llama.cpp can only execute models, you’re missing the best part of the open source ecosystem. It’s a high performance optimization suite. Manually stripping 69% of the VRAM footprint off a brand new model architecture is where real infrastructure value is made. If you want to actually master local inference and deploy models like Google’s massive Gemma 4 12B it on consumer NVIDIA hardware using llama.cpp, you need to learn this pipeline. Let's build it. I just took the raw 22.7 GB Gemma 4 baseline and manually compressed it down to a 7.02 GB Q4_K_M GGUF artifact using llama.cpp. That is a 69% reduction in footprint. No quality loss. No VRAM bottlenecks. Just native, hardware accelerated C++ inference running a full 2,50,000 token context window on a dual NVIDIA Tesla T4 setup. Stop melting your VRAM on unoptimized weights and stop relying on other people's pipelines. Own your stack. I mapped this entire architecture from dynamic binary fetching to raw quantization and real time GPU streaming into a single, bulletproof notebook. Notebook link is in the comments below. Bookmark this blueprint for your next deployment and tell me which quantization works best for your workflow and model.show more

Disappointed with your ICLR paper being rejected? Ten years... ago today, Sergey and I finished training some of the first end-to-end neutral nets for robot control 🤖 We submitted the paper to RSS on January 23, 2015. It was rejected for being "incremental" and "unlikely to have much impact" Our resubmission to NeurIPS was also rejected It now has >4,000 citations (and more importantly, end-to-end training is widely accepted!) It's also cool to think about what's changed and what's the same -- - The network was 92k parameters and trained on ~15 minutes of data - The code was a combination of matlab, caffe, ROS, a custom CUDA kernel for speed, and a low-level 20 Hz controller in C++, all talking to each other. ROS+matlab was as bad as it sounds. - We pre-trained the encoder and did inference off-board on a workstation with a larger GPU. - We were paranoid about varying lighting messing up the network, so we did all the experiments after sunset (so long nights running experiments on the robot past 3 am) Now, we have manipulation policies that are far more dextrous, far more generalizable, and maybe on the cusp of breaking into the real world. :) (the paper:show more

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 🤗!show more

🚨 SHOCKED PASSENGERS WATCH A MAN TURN A COMMERCIAL... FLIGHT INTO AN LED RAVE - AND NO ONE STOPPED HIM Passengers aboard a American Airlines flight from Chicago to Miami say a man stood up mid-flight and started gloving, hand-dancing with flashing LED gloves, and playing music out loud while the plane was cruising at 30,000 feet. With the cabin lights dimmed, the glowing gloves lit up the aisle and reflected across the ceiling and seats. Phones came out almost immediately. Some passengers laughed nervously, others looked stunned, and a few appeared genuinely freaked out watching it unfold while strapped into their seats. What’s throwing people off isn’t just the performance, it’s that there was no announcement and no immediate crew response. For several moments, it simply continued, turning a routine flight into something no one expected and no one asked for. Should this have been stopped immediately or are we normalizing anything goes on planes now?show more

My stream has crashed every stream Ive done for... the last year and a half. I dont know what to do and im about to quit. I have a job- just my hobby is being destroyed ): Here are more videos. Please help. -started with OBS and now a bit less crashes in SLOBS - Nothing in the logs / event viewer says anything is wrong. - Ive reseated my ram. - All gamebar/overlays are off. - no overheating (CPU 38°, Mobo 42°, GPU 53°) -6 month old thermal paste -3 month old bios - all updated firmware on everything/storage/mobo/ram - power hub USB - no plugins - HAGS off - nvidia audio devices off - max game fps set to 60 /playing at low quality - NO overclocking - meld wont work - anker surge protector+2 outlets I cant play games on my powerhouse of a PC, only Playstation is working to avoid lags. Still crashes though. Specs belowshow more

🚨 Alibaba just open sourced a GUI agent that... lives inside your webpage and controls it with natural language. It's called Page Agent and it's not a browser extension. It's pure JavaScript no Python, no Puppeteer, no headless browser, no screenshots. Just one script tag and your web app understands natural language. Here's what it actually does: → Embed it with a single tag or npm install → Control any web interface with plain English commands → Text-based DOM manipulation no OCR, no vision models needed → Bring your own LLM (GPT, Claude, Qwen, anything) → Ships a built-in UI with human-in-the-loop support → Turn 20-click ERP/CRM workflows into one sentence → Optional Chrome extension for multi-tab agent tasks → Works on any web app SaaS, admin panels, internal tools Companies are charging $30/month for AI copilots built on this exact idea. This is 3 lines of code. Your users. Your interface. The AI copilot layer for every web app just got open sourced. 1.6K stars. 100% Open Source. (Link in the comments)show more

Depth Any Video with Scalable Synthetic Data AI physicists... and chemists continue to make strides in depth estimation from video. Check out this new paper featuring some impressive examples. See the thread for more details (unfortunately no code yet). Abstract: Video depth estimation has long been hindered by the scarcity of consistent and scalable ground truth data, leading to inconsistent and unreliable results. In this paper, we introduce Depth Any Video, a model that tackles the challenge through two key innovations. First, we develop a scalable synthetic data pipeline, capturing real-time video depth data from diverse game environments, yielding 40,000 video clips of 5-second duration, each with precise depth annotations. Second, we leverage the powerful priors of generative video diffusion models to handle real-world videos effectively, integrating advanced techniques such as rotary position encoding and flow matching to further enhance flexibility and efficiency. Unlike previous models, which are limited to fixed-length video sequences, our approach introduces a novel mixed-duration training strategy that handles videos of varying lengths and performs robustly across different frame rates 0 - even on single frames. At inference, we propose a depth interpolation method that enables our model to infer high-resolution video depth across sequences of up to 150 frames. Our model outperforms all previous generative depth models in terms of spatial accuracy and temporal consistency.show more

*** SEGA GENESIS/MEGADRIVE - SCALING PART 1 *** I've... been exploring software scaling on the 68k cpu using a background layer rather than scaling sprites. There are a few advantages, no sprite boundaries to worry about so moving pixels is faster also cheaper vertical scaling by adjusting vertical scroll midscreen. The net result is you can manage larger scales more efficiently as the CPU can spend more time working on the horizontal scaling whilst the VDP helps with the vertical scaling (there is still a non trivial cpu cost there though). Here I'm scaling an early image of the Lufthoheit logo - but the goal is to use this in the game also for very large scales / either background effects or bosses. The scaler and Interupt handler are written in Assembly for speed and I think could be faster yet with some optimisations, word alignment moves etc. The scaler has the ability to expand > full screen in height (from a more limited base height) and up to 3x the image width at present. I need to add double buffering and possibly go a bit larger yet. I'm excited to get these effects into the game as large scaling effects we didn't often see on the Genesis ! CYBERDEOUS - Crouzet Laurent Carsten666show more

Google Translate is cooked after this. A developer built... a local AI translation engine that runs 40 languages entirely on your own laptop. It's called LibreTranslate. No API key. No usage limits. No sending your documents to Google's servers. You install it once. It runs forever. Here's what it handles: → Paste text. Translated instantly. → Drop in a file. Outputs the translated version. → Point it at a URL. Returns the page in your language. → Build it into your own app via its local REST API. The speed is not the story. The privacy is. Google Translate reads every sentence you paste into it. Legal contracts. Medical records. Internal emails. Client documents. Every word goes to their servers and stays there. LibreTranslate runs entirely offline. Nothing leaves your machine. Ever. The numbers: → 40 languages supported → Runs on CPU -- no GPU needed → Self-hosted in under 5 minutes → REST API built in for developers → 10K+ stars on GitHub 100% open source. MIT licensed. Price: $0. Google charges nothing for Translate either but it charges you something else. GitHub:show more

Free NVIDIA GPU with 16 GB VRAM GPU for... Running Local LLMs! If you want to master local LLMs but you're waiting until you can afford a $1,500 GPU, you're honestly not going to make it. The open source AI ecosystem is moving way too fast for you to wait on your budget to catch up. Especially when you can build a bleeding edge inference engine from scratch right now, completely for free. You don't need a heavy local rig to start. Google is literally letting you use an enterprise grade NVIDIA Tesla T4 GPU for $0/hour. At standard cloud computing rates (~$0.20/hr), Google Colab’s 4 hour daily free tier hands you roughly $24 worth of data center tier GPU compute every single month. And most people just waste it. Let’s talk about the hardware you get access to for free. The NVIDIA Tesla T4 is an absolute workhorse: - Architecture: NVIDIA Turing (TU104) - VRAM: 16GB GDDR6 (320 GB/s bandwidth) - Compute: 320 Tensor Cores | 2560 CUDA Cores - Performance: 130 TOPS INT8 | 8.1 TFLOPS FP32 - Power: Sipping energy at a max 70W TDP This is the exact same hardware I used to run DeepMind's Gemma 4 26B A4B QAT MoE at a 250,000 context window without a single Out Of Memory (OOM) crash. If you have a web browser and 10 minutes, you have everything you need. I’ve put together a fully documented, cell by cell Google Colab notebook that teaches you exactly how to do this. Here is what the notebook actually teaches you: - How to provision an Ubuntu Linux environment with CUDA 13.0 and verify your driver stack. - How to pull the source code and compile the latest llama.cpp C++ binaries from scratch, specifically optimizing the build for your exact GPU using the -DCMAKE_CUDA_ARCHITECTURES=native flag. - How to directly download quantized local LLMs (GGUF format) straight from HuggingFace using the CLI. - How to manage 16GB VRAM limits, offload neural network layers to the GPU, and push massive context windows. Compile raw llama.cpp, ollama run a model, or spin up the LM Studio CLI. Pick whatever stack you are comfortable with. just start building. No hardware. No credit card. No excuses. Bookmark this post right now so you don't lose the tutorial. Even if you don't have time to run it today, you are going to want this workflow in your engineering toolkit. The link to the free Colab Notebook is in the comments below. Lemme know if you need more tutorials like this.show more

🚨 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 emi) 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.show more

A guy in a kimono made $788,000 on Polymarket... in 3 months without leaving his chair His nickname is gabagool22. A guy in a kimono, sitting in a messy room full of monitors at 3 AM with classical music playing on the TV behind him. The clock on the wall jumps from 02:26 to 06:47 in one timelapse. He's not betting on elections. He's not guessing the next Fed move. He's trading Bitcoin 15-minute binary markets on Polymarket. Profile: 24,525 markets traded. Over $124 million in total volume. Win rate sitting at 99.52%. Profit in roughly 3 months: $788,482. The trick is, he doesn't actually predict anything. He has no idea where Bitcoin is going in the next 15 minutes. He doesn't need to. He just waits for the YES and NO prices on a binary market to drift out of balance, buys the cheap side, hedges with the other, and closes when the spread normalizes. Pure arbitrage. No direction. No emotion. An algorithm watches thousands of micro-markets at once. He just sits there, switches monitors, drinks tea, and lets the bot click for him. One guy in a kimono. No sleep. $788K in the bank.show more

🚨BREAKING: Google just merged Gemini and NotebookLM into one... unified workspace and it changes everything about how you use AI for deep work. It's called Notebooks in Gemini and it's the personal knowledge base that power users have been begging for. You create a notebook for a project, drop in your files, PDFs, and documents, give Gemini custom instructions, and every chat you have stays organized in one place. No more hunting through old conversations. No more re-uploading the same files every session. The wildest part is the sync. Anything you add in Gemini automatically appears in NotebookLM. Anything you add in NotebookLM automatically appears in Gemini. One source of truth. Two powerful apps. Zero friction switching between them. So you can start a research notebook in Gemini, ask it questions all week, then flip to NotebookLM to generate a Cinematic Video Overview from the same material. Next morning, open Gemini and ask it to write a full report on exactly what you just watched. That workflow used to take three apps and a lot of copy-pasting. Now it's one notebook. Rolling out this week to Google AI Ultra, Pro, and Plus subscribers on web. Mobile and free users coming soon. What do you think?show more

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.show more

This is the most hilarious thing I saw and... did today Ran gemma-4-12B-coder-fable5-composer2.5-v1-GGUF locally with 8 GB VRAM at 20+ tok/sec Anthropic's Claude Fable 5 launched June 9. By June 12 it was banned. I can't access it. You can't either. But here's the twist: I'm running a model trained on its chain of thought at 20 tok/s on my RTX 4060 8GB. Locally. Offline. No cloud. No export control. Enter: Gemma4-12B-Coder GGUF (Q4_K_M) Base: Google's gemma-4-12B-it Fine-tuned on verifiable Python CoT data: - Primary: Composer 2.5 real reasoning traces (only passing solutions kept) - Auxiliary: Fable 5 used to redo the hard cases Composer missed. Every training example's reasoning led to code that actually ran. No hallucinated logic. Llama.cpp flags: -m gemma4-coding-Q4_K_M.gguf -cnv -ngl 44 -c 64000 -v (huggingface model link in comments) Flag breakdown: -ngl 44 → offload 44 layers to GPU (tune this for your VRAM) -c 64000 → 64K context window -cnv → conversation/chat mode -v → verbose output The irony writes itself. Anthropic spent weeks telling the world Fable 5 (mythos) is too powerful to release. Then released it. Then got banned from serving it, including their own researchers. Meanwhile: a Gemma 4 12B fine tune, trained on Fable 5's reasoning, runs fully offline on my mid range consumer GPU No API. No cloud. Just me and llama.cpp. This is why local AI matters. Check out the model's link in the comments. How's your experience been with this model?show more

It’s 2025, and it is now okay to sit... in a Police Establishment and signal to world that there is only one god. Police impartiality. How on earth have we got to this place. This gentleman has every right to pursue his faith, I have no issues at all. Pursue it till your hearts content, but not in uniform, and not at work. This small gesture of pointing to the sky will likely offend many who disagree, and cause a complete further breakdown in trust and confidence in policing. What about Jews What about Catholics What about Christians What about Sikhs What about Jehovah Witnesses These aren’t Met Police videos either, these are done on a personal account (open source) on police time for the world to see. This is not on.show more

Microsoft made 100B parameter models run on a single... CPU. bitnet.cpp: The official inference framework for 1-bit LLMs. The math behind 1-bit LLMs is what makes them revolutionary. Traditional LLMs use 16-bit floating point weights. Every parameter is a number like 0.0023847 or -1.4729. When you run inference, you multiply these floats together. Billions of times. That's why you need GPUs, they're optimized for floating point matrix multiplication. BitNet b1.58 uses ternary weights: {-1, 0, 1}. That's not a simplification. That's a fundamental change in the math. When your weights are only -1, 0, or 1: → Multiply by 1 = keep the value → Multiply by -1 = flip the sign → Multiply by 0 = skip entirely Matrix multiplication becomes addition and subtraction. No floating point operations. No GPU required. This is why bitnet.cpp achieves: → 2.37x to 6.17x speedup on x86 CPUs → 1.37x to 5.07x speedup on ARM CPUs → 71.9% to 82.2% energy reduction on x86 → 55.4% to 70.0% energy reduction on ARM The speedups scale with model size. Larger models see bigger gains because there are more operations to simplify. A 100B parameter model running at human reading speed (5-7 tokens/second) on a single CPU. That's not optimization. That's a different paradigm. Why 1.58 bits? Because log₂(3) ≈ 1.58. Three possible values = 1.58 bits of information per weight. The key insight: These models aren't quantized after training. They're trained from scratch with ternary weights. The model learns to work within the constraint. No precision loss. No quality tradeoff.show more