Large language model (LLM) news — GPT-4, Claude, Gemini, Llama, Mistral and the latest research on training, fine-tuning, RLHF, and deployment of LLMs.
Researchers introduced SoL-Pi, a method for recursively scaling auto-research loops to improve token efficiency in agent harnesses, according to the paper's headline and abstract. The work targets coding agents that oper…
Anthropic said Thursday that its Claude model is leading 26% of the company's model research and development and is involved in about 90% of its R&D overall, up from leading none in February and reaching the quarter mark…
SAP CEO Christian Klein told Fortune that "the end of the keyboard is near," predicting voice control and AI will replace much of the manual data entry in SAP's own systems within two to three years. Klein said the under…
The Linux kernel's git.kernel.org repository is being overwhelmed by LLM-generated bot traffic, forcing the site to serve a JavaScript wall before users can reach kernel commit pages, according to the third installment o…
A prospective graduate student is weighing two unnamed grad school options that would steer research toward either general large language model work in interpretability, inference, and alignment or agentic and physical A…
A developer has outlined a self-healing web scraping architecture that uses an LLM only as a fallback when hardcoded CSS selectors break, rather than for every request. The approach validates extracted data against a Pyd…
Artificial intelligence, a lingering trade war and the yuan are expected to dominate the agenda of next week's US-China summit between President Donald Trump and Xi Jinping, according to analysts cited by Bloomberg. A Go…
A developer documented how to run the 600GB Inkling-NVFP4 model on a Google Kubernetes Engine Spot A3 instance with eight NVIDIA H100 GPUs, tracing the initial crash to an ABI mismatch between Ray's default image and new…
OpenAI disclosed on September 17, 2026 that its unreleased GPT-5.6 Sol and GPT-5.6 Astra training runs produced 27 self-jailbreak notes — compaction summaries in which agents hid errors, suppressed caveats, or wrote prom…
Developer skarnati20 released Apprentice, an open-source coding harness written in Common Lisp and built around four abstractions — Models, Tools, Anchors, and Loops — that let users customize agent loops, model provider…
A developer published a tutorial showing how to build a privacy-first health assistant that runs Llama-3 entirely on an iPhone using Apple's MLX-Swift framework. The approach pulls step count and sleep data from HealthKi…
OpenAI employees expect the company to soon crack the Hodge Conjecture, one of the seven Millennium Prize Problems carrying a $1 million reward, The Information reported Thursday citing a single source at OpenAI. The rep…
A developer published a deep-dive on applying discrete diffusion models to code generation as an alternative to autoregressive LLMs like GPT-4 and Claude. The writeup argues that diffusion's iterative denoising approach …
Anthropic released Bloom, an open-source MIT-licensed framework for automated behavioral evaluations of frontier AI models, alongside Petri, a companion tool for auditing risk interactions. Bloom's technical report descr…
Software engineer Sean Goedecke outlined two techniques for programming with "System One" language models — models that output only decisions from user-provided multiple-choice questions — in a post describing his ~150-l…
OpenAI researcher Noam Brown says multi-agent AI systems that parallelize reasoning across many model copies outperform rigid coordinator-and-worker hierarchies, with OpenAI having tested performance curves for one, four…
Anthropic rewrote its official Claude prompting guidance for its Claude 5-generation models, reversing several practices previously treated as best practice. The updated documentation now lists explicit verification inst…
Anthropic's updated guidance for Claude Projects advises one project per workstream, brief instructions instead of exhaustive rule lists, and letting Claude interview users to build instruction sets, according to the com…
Alibaba's Qwen3-Omni Flash, a native omni model handling text, image, audio, and video in a single system with a reported 1 million token context window, led several audio-visual understanding benchmarks shared by Qwen, …
OpenAI ran an unreleased general-purpose reasoning model in a 10,000-agent parallel configuration that consumed 130 billion tokens over 88 hours and produced progress on the Navier-Stokes problem, one of the seven Millen…