Latest publications
Diffusion language models challenge GPT with a 892-token speed record
Researchers have introduced a 100-billion-parameter diffusion model that runs faster than traditional solutions. This opens an alternative path for AI development.
First the brain, then the dumplings: Alibaba DAMO Academy improves robot intelligence
Alibaba researchers urge against rushing to train robots for complex tasks. First, their cognitive abilities and logic need to be optimized.
Noematrix raises funding to build a “universal brain” for robots
Chinese startup Noematrix has closed a Series A round to develop an integrated control platform for next-generation robots. The project aims for full machine autonomy.
Goodbye, latency: Nanyang Technological University speeds up robot response in dynamic environments
Engineers in Singapore have eliminated a critical flaw in VLA models, enabling robots to act without pauses and respond instantly to changes in the real world.
Self-Distillation: How Self-Distillation Will Change AI in 2026
# Self-Distillation: How Self-Distillation Will Change AI in 2026 The artificial intelligence industry has reached a critical juncture.
ProjDevBench: Can AI Build Full-Fledged Software from Scratch?
# ProjDevBench: Can AI Create Full-Fledged Software from Scratch? When we discuss artificial intelligence in software development, we typically recall examples like ChatGPT fixing a bug in a function…
SGLang Ecosystem: Outcomes of a Large-Scale Developer Meeting in Shanghai
# SGLang Ecosystem: How Engineers Learn to Accelerate Neural Networks by Tens of Times Shanghai brought together a small but truly battle-tested group of developers.
IntelliFold 2: new standard in generative scientific AI, now open source
In the world of artificial intelligence oriented toward scientific research, a significant event has occurred: IntelliFold 2 has been presented, a new version of a generative scientific AI platform…
Virtual Pet From Childhood Becomes Reality: Huawei Development
Remember Tamagotchi, those electronic pets from the 90s? It seems the dream of interacting with virtual creatures in the real world is becoming increasingly real.
Call for Submissions for CVPR 2026 Workshop: Security of Multimodal Agents
The artificial intelligence world is developing rapidly, and with it grows the need to ensure the safety and reliability of the systems being created.
This New Year AI doesn't chatter, but pays my bills
As Chinese New Year approaches, when expenses traditionally increase, residents of China are increasingly turning to artificial intelligence not for entertainment, but for help managing their…
Diffusion models: deep analysis and development prospects
Diffusion Models: Deep Analysis and Development Perspectives Diffusion models have become one of the most exciting directions in generative artificial intelligence over the past few years.
New LLM Changes Data Preparation Rules and Leads Hugging Face
In the world of artificial intelligence, data preparation has always been a time-consuming and costly process.
New approach to model drift: Ho Kaiming's work enables generative models to operate without iterative inference
In the world of artificial intelligence, where generative models are becoming increasingly prevalent, the question of efficiency and speed of their operation comes to the forefront.
ChemEval: New benchmark for evaluating chemical large language models
In the era of rapid artificial intelligence development, large language models (LLM) are increasingly applied across various fields, including science.
Pony Alpha: Mysterious Chinese AI Model Sparked Frenzy Online
In recent times, in the foreign internet segment, especially among artificial intelligence specialists, there is active discussion of the emergence of a new large language model (LLM) from China…
ICLR 2026: UIUC Found a Way to Stop LLM 'Overthinking' with One Line of Code
Large Language Models (LLMs), such as GPT-4 and Claude, demonstrate impressive capabilities in text generation, translation, and answering questions.
Waymo and DeepMind Build the Matrix: Why Autonomous Vehicles Need to Imagine Crashes
Imagine you're teaching a child to cross the road. Instead of waiting years for a speeding driver to whiz past or a piano to fall out of a truck, you simply put a virtual reality helmet on him that…
World Models: Why Video Generators Aren't About Cinema, But About the Physics of Reality
When OpenAI released Sora, everyone rushed to discuss how soon Hollywood would be sent to the dustbin of history.
DeepMind and activation function 'mining': why ReLU should retire
For decades, we lived in a world where the ReLU activation function was an unshakeable standard. It was simple as a brick, and effective just enough to avoid getting in the way of neural networks…
DeepSeek vs GPT-5: China's tech market turned into a gladiatorial arena
The era when we evaluated neural networks by their ability to write essays about plastic waste or generate code for a calculator is officially coming to an end.
GPT-5 in a white coat: how OpenAI slashed protein synthesis prices by 40%
While we lazy folks were arguing in the comments about when Sam Altman would deign to release GPT-5 for the masses, the OpenAI team found a far more serious purpose for their new model.
Qwen: Alibaba Burns Three Billion Yuan for Your Tea
While the Western world debates the safety of hypothetical AGI, artificial intelligence networks in China have started buying people tea.
AgentDoG: How a diagnostic collar will tame your AI agents
Do you remember the buzz around AutoGPT and BabyAGI a year ago? It seemed that any moment now we would simply hand an AI agent a credit card, and it would book a vacation, buy groceries, and write…
OpenClaw and 160 Thousand Stars: Why Agent Tools Are a Trap
Listen, 160 thousand stars on GitHub don't just fall from the sky. This isn't simply the success of another repository—it's a collective cry from the industry for help.
Claude Opus 4.6 and GPT-5.3-Codex: Silicon Valley went all-in
Remember when GPT-4's release seemed like the pinnacle of dreams? Forget about it. Today in San Francisco, the air is literally electrified.
Olympic Games in Milan: now with Chinese AI Tongyi Qianwen
Imagine this: Milan, 2026, snow-capped Alps, and Chinese artificial intelligence that knows more about train schedules and ice conditions than any local resident.
OpenClaw: Personal Jarvis on 1% Code Takes Over GitHub
Remember how we dreamed of a personal Jarvis that would solve tasks while we drink our coffee? Reality turned out to be harsher: instead of an intelligent assistant, we got tons of documentation and…
Visual mind: why AI now decides how to think
You've probably noticed how modern neural networks sometimes get stuck on simple problems. They can easily write an essay about Hegel, but sometimes can't understand whether a key is on the left or…
Claude 3.5 Sonnet: Anthropic Deprives Wall Street of Sleep and Billions of Dollars
Imagine you've spent years building a multibillion-dollar empire by selling a convenient interface on top of a database. And then one company releases code that doesn't need your interface anymore.