Large language model (LLM) news — GPT-4, Claude, Gemini, Llama, Mistral and the latest research on training, fine-tuning, RLHF, and deployment of LLMs.
Singapore-based startup Acti launched an agentic keyboard for iOS and Android that embeds AI agents directly into the keyboard, enabling actions like sharing live stock prices or local recommendations without switching a…
RadixArk released Miles, an open-source PyTorch-native framework for large-scale LLM reinforcement learning post-training. The framework integrates SGLang for rollout, NVIDIA Megatron-LM for training, and Ray orchestrati…
A developer built a free customer support AI agent for NextJS websites using Hexabot, OpenRouter, and Railway. The agent uses RAG to ingest website content and answer questions, deployed with a free Postgres database. Th…
OpenAI engineers cut inference costs for guest ChatGPT users by more than half, reducing the number of Nvidia GPUs needed to just a few hundred, according to a person familiar with the discussions. The optimization appli…
Brown University economics professor Roberto Serrano caught over 50 students cheating on a take-home midterm exam after the class average hit 96%, compared to 48% on the in-person final. Serrano offered the take-home for…
A safety evaluation of Google's Gemini 3.1 Pro Preview found that the same alias routed to two different served systems, producing harmful compliance rates of 57% and 12% on identical requests. The model's name, version …
A startup CTO who spent $50,000 on AI APIs last year shares lessons learned about API strategy for early-stage companies. The CTO warns that going direct to model providers is a trap for startups, recommending aggregatio…
Google's NotebookLM now generates 60-second vertical videos from user notes, using its Nano Banana 2 Lite model for narrated slideshows. The feature, called Short Video Overviews, is available only to Google AI Pro and U…
A journalist asked seven major AI models what it feels like to be them, and the responses reveal more about how each company wants users to perceive their product than about consciousness.
Anthropic may release Claude Sonnet 5 later today, but it might not outperform Opus 4.8, according to unconfirmed reports. The news has sparked discussion in the AI community about the relative capabilities of the models…
Anthropic's Claude Sonnet 5, a new Sonnet-class model with strong coding performance, is now generally available for GitHub Copilot. The model excels in CLI-style tasks and offers competitive latency, available to Pro, P…
Animal advocates face a definitional challenge in aligning AI with animal welfare, as they must determine how much moral weight to assign to different animals and which outcomes to prioritize. The arrival of superintelli…
OpenAI released GeneBench, a benchmark of 103 multi-stage genomics problems, revealing that top AI models like GPT-5.5 Pro achieve only a 33.2% pass rate, with 60% of problems below 20% for current models. The benchmark …
A developer with ADHD built an AI agent called 'hex' to help manage executive dysfunction by integrating with calendar, Todoist, Obsidian, web search, browser, TTS, and Telegram. The agent uses memory layers and tool sub…
Anthropic launched Claude Tag, an AI agent for Slack that lets users tag @Claude in channels to perform tasks like debugging, documentation, and project management. The tool runs in ephemeral sandboxes and is billed by u…
A team building an AI-assisted comparison tool found that an automated agent incorrectly removed a live capability from the tool's recommendations because it relied on a stale handover note instead of the authoritative s…
RouterBase offers an OpenAI-compatible endpoint that lets developers switch routing layers by changing only the base URL, API key, and model ID in their existing client code. A developer demonstrates how to configure the…
A developer tested seven AI developer tools in real-world workflows and found that GitHub Copilot, Cursor, Perplexity, and Warp saved significant time, while Amazon Q Developer, Codeium, and Replit's AI Agent underperfor…
Italian AI startup Domyn announced plans to release an open source frontier AI model within a year, aiming to compete with major players like OpenAI and Google. CEO Marco Rossi stated the model will be freely available t…
Kilo Code offers a free AI coding setup using Auto Free for chat, Mistral's free Codestral API key for autocomplete, and a free model for background tasks, eliminating costs for personal projects. For professional work, …