Large language model (LLM) news — GPT-4, Claude, Gemini, Llama, Mistral and the latest research on training, fine-tuning, RLHF, and deployment of LLMs.
A new paper argues that if large language models are considered sentient, then by the same metrics the classic video game Age of Empires II should also be considered sentient. The discussion was featured on a podcast tha…
NVIDIA has released a GenAI LLM Certification Lab that guides developers through building a production-ready fine-tuning and optimization pipeline. The lab covers data preparation, LoRA fine-tuning with 4-bit quantizatio…
Engineers are observing that larger context windows in AI models degrade output quality, a phenomenon called 'context rot.' As context grows, attention mechanisms become diluted, leading to vaguer and less precise respon…
A developer with over two decades of experience in technology describes how artificial intelligence has become an invisible infrastructure in daily life, citing practical use cases such as LLM-based productivity assistan…
OpenAI deployment chief Arnaud Fournier said in an interview that the company aims to embed AI deeply within large corporations using its own engineers, citing explosive growth of Codex and a sharp drop in AI prices. He …
Google released the new Google Home Speaker, its first smart speaker in six years, featuring Google Gemini instead of Google Assistant. The device offers smart-home control, stereo pairing, and a more human-like assistan…
Researchers at UC Berkeley introduced LEVI, an AI-driven framework that optimizes systems algorithms at 3-7 times lower cost than existing methods. By using smaller, cheaper models for most mutations and reserving larger…
A developer built a tokenization guardrail for AI agents that prevents sensitive data from reaching large language models. The system replaces real customer data with opaque tokens, swaps them back only during tool execu…
A family of five in the UK openly uses AI tools like Claude in their daily lives, integrating voice assistants into work, meal planning, and parenting. The parents teach their three children aged 5 to 12 to use AI as a t…
A developer achieved 3x better batch throughput with Qwen3.5 by fixing vLLM on AMD's Strix Halo using the Lemonade Server, enabling more efficient AI inference on AMD hardware.
Chinese AI lab Zhipu is considering a multibillion-dollar share sale in Hong Kong after its stock surged roughly 2,000% since January, pushing its market value above HK$1 trillion. The rally follows the launch of its ope…
Alibaba Group slashed prices for its Qwen AI models by up to 80% on its Qoder coding platform, targeting US workday hours to attract global developers amid fierce competition with US and Chinese rivals like Anthropic and…
Claude Code can improve other AI agents without training data in four of seven tested applications, performing as well as with data. Data helps only where Claude Code's prior knowledge of the task runs out, and the drift…
AI companies like Anthropic are building powerful models without adequate regulation, echoing J. Robert Oppenheimer's 'technically sweet' justification for developing the hydrogen bomb. Anthropic co-founder Jack Clark ar…
Noah Lindner and the team at Bevyl launched Stupify, an anti-slop code review tool for AI agents that judges code against a user-defined taste corpus, running on OpenAI's Codex via exe.dev. The tool aims to improve code …
A developer built an end-to-end AI-SRE system using HolmesGPT, mirrord, and a Claude wrapper to autonomously verify fixes on a real GKE cluster. The system investigated a checkout error, generated a code patch, and teste…
A B2B SaaS platform discovered that two of its six AI features were consuming 71% of its $4,200 monthly AI budget after implementing per-feature cost attribution. The Compliance Checker and Audit Trail Narrator were sile…
CoinGecko launched Gecko AI, a conversational research tool that lets premium subscribers query its crypto datasets using natural language. The tool, released on April 22, 2026, aims to differentiate CoinGecko as a resea…
GLM 5.2, an open-weight coding model from Z.AI, outperformed Claude Opus in four real-world coding tasks including a codebase audit, UI redesign, and autonomous bug hunting, costing only $3.36 for 6 million tokens. The m…
Alibaba's Qwen team released Qwen-AgentWorld, a language model that simulates complex environments natively, replacing external simulators for training AI agents. The model, trained on over 10 million trajectories across…