#LLM
Curated coverage on «LLM»: releases, research, benchmarks, open-source.
DeepSeek V4 Pro vs. Claude Sonnet 4.6 on 50 tasks: where it saves money and where it adds risk
A 50-task test built from a Russian developer's real workload shows DeepSeek V4 is far cheaper than Claude Sonnet 4.6 but errs more often on calculations, OCR and…
China Moves to Block Meta's Manus AI Purchase: Can Meta Save the $2 Billion Deal?
China has demanded that Meta unwind its already completed Manus AI acquisition, showing that the AI agent race turns on control of technology as much as money and…
T-Technologies finds a way to curb sycophancy in GPT and DeepSeek without retraining
T-Technologies' R&D center shows how to test and reduce the tendency of GPT, DeepSeek and other LLMs to agree with a mistaken user, without full retraining.
Developer teaches n0x browser agent to open websites and take screenshots
OpenAI dismisses sales growth concerns, says business demand is strong
Bloomberg rebuilds Terminal around AI: the ASKB chat will become its main interface
Anthropic's Claude Mythos Raises the Threat: AI Puts Cyberattacks Within Reach of Novices
Anthropic is testing Claude Mythos: a leak reveals a model with a reported 10 trillion parameters
An internal Anthropic leak has exposed Claude Mythos, which the company considers its most powerful AI and is not yet ready to release publicly, citing cost and…
Anthropic accidentally exposes Claude Code source, raising fresh security questions
Anthropic mistakenly shipped a Claude Code update pointing to an archive of nearly 2,000 internal files, and the leak spread across GitHub and X within hours.
Anthropic bets on Cowork: the agent could outgrow Claude Code in market size
How a Single System Prompt Turns an LLM into a Reliable Tool: Tests on Qwen and DeepSeek
T-Technologies Group on open source in AI/ML: how LLM work is organized inside the company
VectifyAI's PageIndex offers embedding-free retrieval for long documents
LLM system cuts control-card preparation at a steel plant from 2 hours to 5 minutes
At a metallurgical plant, an LLM system now builds control cards in 3–5 minutes instead of two hours. What worked was not a universal prompt but a set of rules tied…
Habr AI shows how to add memory and context to a Python LLM chat with Ollama and LiteLLM
The latest part of a Python chat tutorial with Ollama and LiteLLM shows how to store message history, pass context to the model and cap memory without complicating…