All news
7223
Russian scientists created a monitoring system for smoke and dust that is 20 times cheaper than a thermal imager
Russian researchers have presented an AI video surveillance system that detects a moving person in smoke and dust, can be deployed in a coup
Tokentap and MitM Proxies for LLMs: How to Monitor Tokens, Costs, and Data Leaks
Tools like Tokentap make it possible to place a MitM proxy between an LLM CLI and a cloud model to track token usage in real time and reduce
PrimeCode adds a full graphical interface to OpenCode CLI right inside VS Code
The PrimeCode extension turns OpenCode CLI into a full interface inside VS Code: it shows tool calls, file changes, context, and lets users
Project Nomad turns a Debian PC into an offline knowledge hub with local AI, maps, and Wikipedia
OpenIDE adds ACP: how the JetBrains and Zed protocol changes how AI agents work in IDEs
Owlcat Games uses generative AI in The Expanse: Osiris Reborn, but the release version will be made by humans
DeepSeek and Qwen tried to get through "Everlasting Summer," but got stuck in the labyrinth
Google introduces TurboQuant: how the new compression lowers the cost of local AI
Google’s new algorithm compresses the KV cache of language models in real time, promising up to 6x memory savings and faster inference, but
Noi brings ChatGPT, Claude, and Gemini into one desktop app without extra switching
Noi combines popular AI services and local models in one desktop interface so users can work with ChatGPT, Claude, Gemini, and Ollama withou
Russian scientists propose cooling for AI data centers that cuts electricity use by up to 22%
OpenClaw turned out to be a costly autonomous agent: 81 million tokens and server conflicts
Meta released Tribe v2 — a model that predicts the brain's response to video, audio, and text
Google launches Gemini 3.1 Flash Live — voice AI that is increasingly hard to distinguish from a human
Nvidia Nemotron-Cascade-2 was run at home on a GeForce RTX 3090 at up to 150 tokens/s
A local LLM enthusiast tested Nemotron-Cascade-2 on a home PC with a GeForce RTX 3090 and showed that the 30B model can work as a fast assis
TAPe reached 74% accuracy on COCO and began moving away from standard transformers
The authors of the TAPe diary said that the model, trained on entirely synthetic data, reached 74% classification accuracy on COCO and is no