All news
10835
Claude (Anthropic): Memory and Neural Network Architecture
The author of the Aurora project describes a late-night conversation with Claude about AI memory, identity and the architecture of a digital entity meant to…

How to Fine-Tune a Search Model: NVIDIA's One-Day Method
NVIDIA and Hugging Face have published a step-by-step recipe that turns a base embedding model into a specialized search tool for internal documents in a few hours.

deepvk vs OpenAI text models: a benchmark on court rulings
The author compared 7 embedding models and 4 rerankers on 858 intellectual property cases and found that the local deepvk USER2-base performs almost on par with…

Palantir Deepens Its UK Footprint: FCA Contract Opens Access to Sensitive Data

How to Use the Qwen AI Model for Precise Answers

AI adoption: why 90% of business pilots never reach production

SK Hynix to spend nearly $8 billion on ASML EUV tools for next-generation memory

Sber's GigaChat: A Free Chatbot Alternative to ChatGPT, With Paid Features Ahead
Sber wants GigaChat to replace dozens of apps as a universal AI assistant, and to eventually charge for its most valuable features, chiefly agents and B2B services.

AI in Energy: How Data Centers Are Reshaping Electricity Demand
Energy Vault has bought a 175 MW battery project near Dallas, Texas, betting on fast-growing power use by AI data centers and a shortage of firm capacity on the…

Sberbank: Russia Needs Sovereign AI, but It Can't Do Without Controlled Foreign Data

Yandex expands robot delivery to the Moscow region, new parts of Kazan, and Nizhny Novgorod

AI Photo Editing: The Otredach Case on VK

AI for document recognition: Smart Engines secures a patent

Arm AGI CPU: the first processor built for neural networks and agentic AI
Arm has moved beyond licensing for the first time and unveiled its own server chip, the AGI CPU: 136 cores, a focus on agentic AI and a claim of more than double…

JAIST and Princeton Researchers Build NTAC, an Algorithm That Types Neurons by Their Connections
A JAIST and Princeton team shows that NTAC identifies neuron types by synaptic connections rather than shape, reaching over 90% accuracy in minutes on an ordinary…









