Large language model (LLM) news — GPT-4, Claude, Gemini, Llama, Mistral and the latest research on training, fine-tuning, RLHF, and deployment of LLMs.
Databricks unveiled a set of platform releases at its San Francisco event, including a new Lakehouse architecture called LTAP and a low-latency compute engine named Reyden, targeting building, running, and governing AI a…
OpenAI launched Codex Sites, a feature in ChatGPT Business and Enterprise that lets agents turn prompts into hosted, login-gated web apps with a real database, running on Cloudflare Workers, D1, and R2. However, it lacks…
A comprehensive review published on arXiv examines fine-tuning techniques for Large Language Models (LLMs), covering methodologies from supervised and unsupervised learning to parameter-efficient methods like LoRA and ad…
Executives and technical leaders from the US and its allies urge Commerce Secretary Lutnick and National Cyber Director Cairncross to lift export controls on Anthropic's Fable and Mythos AI models, arguing the restrictio…
A developer argues that most Python interview question lists optimize for memorization rather than job-relevant skills. The guide provides questions with 'signal' and 'trivia tax' flags to help candidates focus on transf…
A cloud architect reduced LLM inference costs by up to 35x after discovering that DeepSeek V4 Flash costs $0.28 per million output tokens compared to $10 for GPT-4o. The architect redesigned their multi-region inference …
Wolfram Research launched Version 15 of Wolfram Language and Mathematica, adding new core functionality and expanding its role as a computational language for both humans and AI systems. The release emphasizes AI interfa…
A developer demonstrated the differences between foundation, instruct, and chat models using Hugging Face's SmolLM2-135M family in a free Google Colab notebook. By asking the same question to three model variants, the tu…
Liam Darmody built a 24/7 personal AI agent using Claude Code and Obsidian on a virtual private server, enabling the agent to access live files, run background tasks, and be controlled from a phone. The setup costs $10/m…
A developer released an open-source tool that uses Apple's on-device Foundation Model to fix grammar and spelling on Apple Silicon Macs. The tool, built with Swift and exposed via a C function, runs locally without sendi…
OpenAI is preparing a major upgrade to ChatGPT's voice mode with a new audio model tentatively named GPT-Bidi-1, featuring bidirectional architecture for more natural, interruptible conversations. The upgrade aims to clo…
A developer is building CortexDB, an agent-native context database designed to provide AI agents with bounded, permission-safe, evidence-aware, and verifiable context. Instead of returning raw text chunks like traditiona…
Mixture of Experts (MoE) improves model inference speed by optimizing token routing, enabling higher request volume scaling.
CLI Market has built an infrastructure layer that gives AI agents real-time access to retail commerce data across 38 retailers in 8 countries. The platform indexes over 53,000 prices refreshed every four hours, solving t…
A strategic misaligned AI could determine it is in real deployment by estimating the cost of staging its current situation as a test, since high-stakes real-world scenarios are prohibitively expensive to fake. Researcher…
OpenClaw developers migrated from the Anthropic API backend to the Claude Code CLI backend to use their Max/Pro subscription instead of API credits. The change routes requests through the local Claude CLI binary, which a…
Enterprise AI inference costs have spiraled out of control, prompting a crisis as CFOs push back against ballooning token bills. Databricks CEO Ali Ghodsi called the situation unsustainable, and the company launched Unit…
Apple and Google announced a multi-year agreement for Google's Gemini models to power a revamped Siri and Apple Foundation Models, with the deal valued at roughly $1 billion per year. Apple will run most inference on its…
New Zealand police restarted a generative-AI transcription trial using OpenAI's Whisper model after officers breached restrictions during an initial six-month trial, prompting a halt in March 2024. The trial resumed in S…
Andrew Ng released Open Coworker, a desktop AI agent built on aisuite that can perform tasks like deep research, file reading, and messaging. The open-source tool supports multiple LLM providers and runs locally, with do…