obsidian-mcp-server v3.2.0 adds Omnisearch mode and cursor pagination.
Headroom is a local-first compression layer that reduces AI agent input by 60-95% while being able to restore the original.
The Gemini app is expanding its reasoning level options and third-party app integrations.
It summarizes how to design agent evaluation harnesses that go beyond static benchmarks.
Cursor has released Composer 2.5, a coding agent built with reinforcement learning, synthetic data, and new distributed training techniques.
Anthropic has acquired Stainless to strengthen Claude's SDK and MCP connectivity.
llama.cpp b9200 includes an optimization that improves MTP prompt processing speed.
Musinsa automated everything from slow query analysis to PR reflection with query-engineer.
Across 7 experiments, people judged AI responses as more confident than human responses.
Content creators are using tarpit tools that poison LLM training data in response to unauthorized scraping.
This introduces research that models the morphogenesis and regeneration processes of living organisms through Neural Cellular Automata.
Haystack has launched to analyze AI-generated Pull Requests and support automatic classification and review.
OpenAI and the Malta government are launching a program to give one year of free ChatGPT Plus to those who complete an AI education course.
Skales, a local AI agent connecting directly to Ollama and optimized for privacy protection and low-spec environments, has been released.
A new memory retrieval system achieved SOTA on the LongMemEval benchmark using Gemini Flash.
Publicis has acquired LiveRamp for $2.5 billion to strengthen its agentic AI data capabilities.
This is research analyzing the differences in social behavior patterns shown by different LLM agents in a virtual town.
The UK Government Digital Service (GDS) has recommended maintaining an 'open by default' policy for AI and code security in the public sector.
ROCm 7.13 nightly adds Strix Halo optimizations and open-sources ROCprof.
An update has been released for llama.cpp that optimizes prompt decoding performance when using MTP.
Semble has been released, reducing token usage by 98% and speeding up search for agents' code search.
MiroThinker-1.7, an open-weight deep research agent based on Qwen3 MoE, has been released.
OpenRouter costs about 1/3 as much as Apple Silicon and is 2x faster, based on comparable models.
Anthropic has released an official free training course for learning Claude's Tool Use feature.