Technical AI news, automatically curated and generated
AMD has announced the acquisition of AI chip startup Taalas, a strategic move designed to enhance AI inference performance by integrating specialized models directly into silicon hardware. While the specific financial te…
→ View original source
Researchers have introduced Activity Frames, a deterministic, zero-model compilation pipeline designed to transform passive screen-capture streams into structured memory for computer-use agents. Unlike traditional method…
→ View original source
Researchers have introduced MameLoshnLM, the first open-source 8B-parameter language model specifically engineered for the Yiddish language. To overcome the limitations of noisy, machine-translated multilingual datasets,…
→ View original source
The researchers introduce AgentOPSD, a critic-free, recursive framework designed to improve turn-level credit assignment in long-horizon, multi-turn agentic reinforcement learning tasks. Addressing the limitations of tra…
→ View original source
EnvACE introduces a novel agentic reinforcement learning framework designed to mitigate the high costs and scaling limitations associated with external environment interactions during LLM agent training. Instead of relyi…
→ View original source
Castform has developed a platform designed to enable Reinforcement Learning (RL) post-training for open-source models, allowing them to match or outperform frontier models like gpt-5.6-sol in agentic retrieval tasks at a…
→ View original source
OpenAI has overhauled ChatGPT's voice infrastructure by introducing GPT-Live, a full-duplex architecture designed to eliminate the latency bottlenecks inherent in traditional turn-based voice-activity detection. Unlike p…
→ View original source
OpenAI has implemented a significant update to its GPT-5.6 Sol model, focusing on enhancing multi-step reasoning and instruction-following capabilities. This technical refinement was achieved through improved training da…
→ View original sourceCherry Studio is a cross-platform desktop application designed for Windows, macOS, and Linux that provides a unified interface for diverse Large Language Model (LLM) providers. The software facilitates multi-model simult…
→ View original source
An analysis of over 40,000 game runs involving human-in-the-loop AI agent command approvals reveals significant security vulnerabilities in manual oversight. The study found that players missed an average of 1 in 3 threa…
→ View original source
To address the fragmentation of AI agent integrations, OpenAI, Microsoft, Amazon, Vercel, and Anysphere have released "Agent Plugins," an open-standard specification version 1.0 designed to enable cross-platform interope…
→ View original source
Anthropic's Claude Mythos LLM has demonstrated new cryptanalytic capabilities, specifically targeting post-quantum and symmetric primitives. The model successfully identified a key-recovery attack on the HAWK post-quantu…
→ View original source