But the most interesting thing about Anthropic's IPO data is that they raised a total of $141.1 billion, and still have $131.7 billion in their accounts. It turns out that throughout its history, Anthropic has spent about $9.4 billion, while in just 3 years its revenue has grown from zero to $16.3 billion in...
Telegram — blockchainRF (блокчейн / мозг / space / WEB 3.0)
🤖 Sam Altman explained how artificial intelligence can launch an endless cycle of technological development. The head of OpenAI believes that the future of AI is not only about creating models that independently improve their own intelligence. There is another mechanism...
Every fourth skill for AI agents turned out to be unsafe - NVIDIA introduced a tool to check them. NVIDIA has released a free SkillSpector scanner that analyzes skills that enhance the capabilities of AI agents. Such skills allow neural networks...
Tencent releases Hy-Nomas: open-source interactive video world model that runs on a single RTX 4090 🌍 The model builds the world from video segments, and you control it with text and camera trajectory: at each step you can change the scene, style, action, or angle, and the world's memory…
🤖 NVIDIA has released Isaac Lab 3.0, a platform for training robots in the virtual world. The company has introduced a major update to Isaac Lab, which allows robots to be taught complex movements and actions in simulation before transferring the acquired skills to the real world...
3Blue1Brown style math animations are created using Manim. They look really impressive, but rendering even one video takes quite a long time. ManimGX retains the familiar Python syntax of Manim Community, but uses Rust and GPU to...
Paragon is in touch - summing up the results of the day! The main thing for the day🟠22.10 🟠Claude Duet - AI agent writes to you first 🟠Tesla - Optimus has gone into production 🟠Perplexity Computer - runs entirely on Mac 🟠PHP - two dangerous holes have been closed 🟠Meta Muse - 10 million downloads in 40 days 🟠Bark 3.0...
📌 Keynote GTC Berlin 2026 The main topics are physical AI, sovereign computing and the transformation of data centers into the format of factories that produce intelligence as a product. Over the year, Nvidia invested almost $10 billion in Europe, announced 14 investments in startups and counted more than 12...
🌟 CLM-8B: an open and fast model for selecting agent actions Stanford, in collaboration with NVIDIA, has made the first model of the Contrastive Language Models class publicly available. This is an open analogue of Jev, which was presented on September 15 by the TypeSafe AI startup of a former researcher...
✔️ Dream Labs presented a model for controlling unfamiliar robots Four former NVIDIA researchers from the GR00T robotics project team came out of stealth with their startup and opened the public API of the Dream-1 model. The company was led by Joel Jang, who led the GR00T…
NVIDIA вместе с Нобелевским лауреатом выпустили ИИ-модель для дизайна биомолекул, которая в вычислительных тестах обходит конкурентов по доле успеха и скорости Команда Дэвида Бейкера и Nvidia опубликовали веса и код своей ИИ-модели RFD4-Proteina. Модель работает в общем…
Telegram — blockchainRF (блокчейн / мозг / space / WEB 3.0)
If you have two or three Macs at home and want to run a large AI model locally, the memory on one computer may not be enough. exo allows you to combine multiple devices into one AI cluster. A model that does not fit into the memory of one computer can be distributed among several...
💸 Anthropic может потратить более $750 млрд на облачные мощности — но их всё равно не хватает По данным The Wall Street Journal, создатель Claude столкнулся с нехваткой вычислительных ресурсов у AWS, Google и Microsoft. При этом облачные гиганты всё чаще воспринимают Anthropic…
✔️ DeepSeek и Huawei вместе делают следующее поколение ускорителей Ascend Инженеры Huawei работают прямо в офисе DeepSeek, а заняты команды оптимизацией интерконекта и распределением вычислительных графов на уровне железа. Раньше DeepSeek от обучения модели на Ascend отказалась…
DeepSeek has open-sourced DeepInfer, the inference engine it is using to launch V4.1-Flash at scale. https://github.com/deepseek-ai/DeepInfer It is this code that helps to quickly service requests to models and reduce the cost of their operation. For companies…
DeepSeek has opened the source code of DeepInfer, its own engine for running AI models. It is this that ensures DeepSeek-V4.1-Flash operates with high performance and minimal latency. The main thing: - Speed up inference - more requests in less time. - Effective...
✔️ Nvidia transfers GB202 chips to the professional segment According to sources, the real reason for the shortage of RTX 5090 video cards is the cannibalization of GB202 chips - the company is redistributing crystals in favor of the professional workstations segment. The card has already disappeared from...
✔️ NVIDIA has released the multimodal model Nemotron 3.5 Super VL The newest member of the Nemotron family with open scales understands text, images and video up to two minutes long. Under the hood, MoE has 120 billion general and 12 billion active parameters. NVIDIA is counting on agents who...
Nvidia has an open world model with 2.6 billion parameters. It lets you turn an image, a text prompt, and a camera motion trajectory into a controllable virtual world. All on a single GPU. - Code available on GitHub. - Research paper published on arXiv. - Suitable for…
📺 ИИ научился самостоятельно распространяться по умным устройствам Исследователи Anthropic проверили, сможет ли модель копировать себя между устройствами без доступа к дата-центрам. В тестовой сети ИИ запустили на 15 недорогих ТВ-приставках. Модель сама находила другие…
✔️ Science Corporation впервые имплантировала зрительный чип PRIMA Биотех-стартап бывшего сооснователя Neuralink Макса Ходака расширил сферу применения своего интерфейса мозг-компьютер на неизлечимое наследственное заболевание Штаргардта, приводящее к прогрессирующей слепоте…
Anthropic: very soon small AI models will be able to self-replicate on smart ovens, TV boxes, etc. without access to data centers The authors studied autonomous replication and adaptation (ARA) - this is when the model itself deploys and distributes without human intervention,…
Telegram — blockchainRF (блокчейн / мозг / space / WEB 3.0)
⚡️ Microsoft is betting on its chips: Maia will become the company's main inference platform From 2027, the company will scale the Maia family to multi-gigawatt deployments, and Maia will become the main inference system for its own products and services. At the same time...
🚀 Qwen introduced QwenGyre: training AI agents up to 1.85 times faster Modern agents can perform tasks for hours, generating hundreds of thousands of tokens. When training such systems, GPUs are often idle, and repeated trajectories waste computing resources. QwenGyre…
🖥 В России предложили разрешить зарубежные GPU для критически важных систем Правительственная рабочая группа предложила временно разрешить использовать иностранные графические ускорители и ИИ-решения на объектах критической информационной инфраструктуры. Причина — нехватка…
Paragon is in touch - let's start the day! The main thing for the day🟠12.10 ⠀ 🟠Claude - dangerous advertising on Google 🟠Android - the virus is already in the firmware 🟠Microsoft Decision-1 - faster than GPT-6 Sol 🟠Anthropic - the agent sent the real form 🟠ChatGPT - without it people give up 🟠Yandex is the third data center...
✔️ Google открыла публичный доступ к SynthID Detector Компания запустила веб-версию инструмента SynthID Detector, позволяющего анализировать изображения, видео и аудио на наличие невидимых меток, внедряемых ИИ на этапе генерации. Ранее утилита тестировалась в закрытом режиме…
Восстановить работу сервисов после одновременного повреждения нескольких крупных дата-центров гораздо сложнее, чем просто переключить нагрузку на резервные серверы. Инфраструктура Яндекса обслуживает миллионы пользователей, а её компоненты тесно связаны между собой. Когда сразу…
✔️ Криптоиндустрии пора готовиться к взлому кошельков с помощью ИИ Исследователь Джастин Дрейк предложил блокчейн-индустрии готовиться к режиму бункера, то есть к массовому переводу средств на новые адреса на случай, если ИИ поможет взломать ECDSA. На этой схеме цифровой подписи…
Mistral обучила мощнейшую ИИ-модель на 3800 чипах NVIDIA Blackwell — и показала дата-центр, похожий на вокзал 💀 В интернете угарают с фотографии французского дата-центра, которой поделился сооснователь Mistral. По его словам, новую модель обучали на 3800 мощнейших чипах NVIDIA…