Large language model (LLM) news — GPT-4, Claude, Gemini, Llama, Mistral and the latest research on training, fine-tuning, RLHF, and deployment of LLMs.
Google DeepMind released DiffusionGemma, an open-source Apache 2.0-licensed diffusion-based large language model that generates text at over 1,000 tokens per second on a single H100 GPU, achieving up to 4x faster through…
Vigneshwaran S. released Vilvona AI, a self-hosted AI assistant that connects to messaging and productivity apps including WhatsApp, Telegram, Gmail, and Slack while keeping all data local. The tool features a Tamil and …
A developer has identified a persistent security threat to AI agents called memory poisoning, where malicious instructions stored in an agent's memory can influence all future interactions indefinitely. The developer con…
Kimi.ai released Kimi-K2.7-Code on Friday, an open-source coding model available via Hugging Face with its weights and code. The company claims the model, positioned specifically for coding and agent tasks, improves over…
A new benchmark called FactWorld reveals that hybrid AI models combining transformer and recurrent architectures can simultaneously excel at both associative recall and state tracking, capabilities that pure transformer …
A developer has created a 60-second audit script to measure the "MCP server token tax"—the context budget consumed by tool definitions before an agent performs any useful work. Running the script against the real filesys…
A developer has published a recipe for writing effective Flutter agent skills, synthesizing official guidance from Flutter, Anthropic, Google, and OpenAI into a single copy-pasteable format. The recipe emphasizes that a …
Apple's new AI-powered dictation feature in iOS 27 beta 1 is disabled by default on iPhone 17 Pro and iPhone Air, requiring users to manually enable it in Settings under General > Keyboards. The on-device "Advanced Dicta…
Google's DiffusionGemma, an experimental open model using discrete diffusion for text generation, offers a parallel approach that can outperform token-by-token LLMs in throughput-sensitive workloads. Developers should ev…
An open source developer has announced he will no longer accept unsolicited pull requests after determining that nearly all contributions to his projects are now generated by large language models. The developer, who pre…
MiniMax released its M3 multimodal model on NVIDIA's accelerated infrastructure, offering a free public endpoint via NVIDIA's API catalog. The 428-billion-parameter model processes text, images, and video with a one-mill…
International Monetary Fund Managing Director Kristalina Georgieva warned Wednesday that advanced AI models like Anthropic's Mythos "can be used to destroy the financial system" if weaponized by malicious actors. Speakin…
French AI lab Mistral AI is in early discussions to raise approximately €3 billion ($3.5 billion) in a funding round that would value the company at around €20 billion ($23.15 billion), nearly double its valuation from l…
A developer has demonstrated how to build structured knowledge graphs from unstructured documents using Google's Gemini AI model. The approach involves prototyping with Gemini to extract relationships and entities from r…
A developer successfully configured a local coding agent on macOS using Gemma 4 26B-A4B and Qwen3.6 35B-A3B models with llama.cpp, achieving 72.2 tokens per second generation speed through MTP speculative decoding. The s…
A developer detailed three methods for reducing LLM inference costs without sacrificing output quality. The approach targets redundant context in every request, over-retrieval of irrelevant chunks, and unnecessarily long…
A data engineer replaced 1,000 brittle rules with three AI calls in a hybrid data quality framework on Databricks SQL, using deterministic checks for speed and AI for semantic judgment to audit messy vendor data. The app…
Surplus, a new incubator organized by Manifund, is accepting applications through June 24 to fund for-profit software startups focused on AI safety, effective altruism, and public goods. The program targets founders buil…
A developer built an API for sales research that finds LinkedIn profiles with a 75% success rate and returns decision-maker information, but is struggling with AI agents like Claude losing context of the API key, endpoin…
Google DeepMind researchers developed a model diffing agent that automatically discovers and validates behavioral differences between two large language models, addressing the limitation of standard evaluations that can …