Large language model (LLM) news — GPT-4, Claude, Gemini, Llama, Mistral and the latest research on training, fine-tuning, RLHF, and deployment of LLMs.
Llama.cpp developer released ggrun, an auto-tuning tool that measures GPU, RAM, and PCIe topology to compute optimal multi-GPU and MoE expert placement for GGUF models, serving an OpenAI-compatible API. Benchmarks show g…
The Trump administration has asked OpenAI to stagger the release of its upcoming model, GPT-5.6, initially releasing it only to a short list of trusted partners with customer-by-customer government approval. CEO Sam Altm…
A GitHub repository called design.md, which provides AI agents with a persistent design document, has gained over 1,400 stars. The approach addresses context fragmentation in agent development by offering a single source…
AI search engine Exa raised $250 million in Series C funding at a $2.2 billion valuation, led by a16z, to power AI agents with high-quality web search. Exa already serves over 5,000 companies including Cursor, Cognition,…
Local open-weight large language models (LLMs) with 30 billion parameters and a mixture-of-experts architecture achieve roughly 40 tokens per second on a Mac or DGX Spark, matching GPT 5.5 Pro subscription speeds and sol…
A developer explains that Cursor AI is more than an AI-powered code editor, detailing its features including Rules, Skills, Hooks, Plugins, MCP, Automation, and Customization. The guide uses real-world examples to show h…
A developer building a Catholic theology AI app implemented a groundedness evaluation pipeline to prevent hallucinations and ensure doctrinal accuracy. The app, Catholic Theology: AI & Faith, combines a chatbot with dail…
NVIDIA CEO Jensen Huang and Anthropic's Claude Code chief Boris Cherny have both declared that the age of prompting AI models is over, pushing instead for 'loops'—LLMs prompting other LLMs continuously to burn more token…
OpenAI is reportedly considering delaying its IPO from 2026 to 2027 due to a tech selloff that has depressed AI stock valuations, with CEO Sam Altman insisting on a $1 trillion minimum valuation. The company's confidenti…
A developer building Naturalmelo, an AI text detector, discovered that the hardest challenge was not training the machine learning model but understanding user expectations. Users cared less about accuracy metrics and mo…
A developer reported saving 5-6 hours of QA time by using an AI-assisted workflow to generate unit tests for a 23-function module, completing a first draft in under 90 minutes. The same approach compressed debugging time…
AI now writes nearly all code at startups like Alma, Chainguard, and Wordsmith AI, with Anthropic's Claude Code as the dominant tool. Founders report massive speed gains but also warn of a 'cleanup tax' from bugs and slo…
Payload CMS v4's plugin now exposes every collection as an MCP tool with full CRUD by default, requiring opt-out for sensitive operations. Developers must review collections after upgrading to prevent accidental exposure…
A developer built a free technical checker that evaluates whether a website is ready to be crawled, understood, extracted, and trusted by AI-powered search engines. The tool separates 'AI visibility' from 'AI readiness,'…
A developer argues that AGENTS.md should function as a bootloader rather than a knowledge base, advocating for staged context delivery via the Model Context Protocol (MCP). The developer explains that MCP can serve as a …
A developer outlines a phased approach to onboarding AI coding agents onto existing projects, arguing that poor agent performance often stems from insufficient project documentation rather than model limitations. The gui…
Claude Opus 4.8 successfully translated 31 of 33 pandas scripts to idiomatic Polars code in one-shot tests, enabling faster data pipelines on cheaper hardware. The translations ran correctly and output matched, though tw…
NeuralWatt, a US-based AI inference provider, introduced energy-based metering for LLM inference, charging by kilowatt-hour instead of per token. A user reported an average 82.9% cost reduction compared to token-based bi…
OpenAI's GPT-5.6 rollout now requires U.S. government approval on a customer-by-customer basis, following a request from the government. CEO Sam Altman stated this is not a preferred long-term model, as AI labs fear a de…
At AI Native DevCon London, engineering leaders from various organizations discussed the shift from individual AI coding tools to enterprise-wide agent governance. The roundtables highlighted that while initial adoption …