Large language model (LLM) news — GPT-4, Claude, Gemini, Llama, Mistral and the latest research on training, fine-tuning, RLHF, and deployment of LLMs.
A developer at Mattrx, a multi-tenant marketing-analytics SaaS built on .NET 9 and Azure, reports that adopting the Model Context Protocol (MCP) eliminated 14 point-to-point integrations by replacing them with 3 MCP serv…
A developer built an interactive 2D retro action-roguelike game called GPU Survivors that simulates the architectural limits, failure modes, and optimization hyperparameters of running a Large Language Model under load. …
A new technical report details the sharded inference of a 229-billion-parameter mixture-of-experts model across five consumer GPUs in five countries over the public internet, achieving 12.6 tokens per second interactivel…
On 20 March 2026, WordPress.com enabled AI agents to autonomously write, edit, publish, and manage entire websites via the Model Context Protocol, adding 19 new writing capabilities across six content types. The platform…
A new visual knowledge canvas tool for learners integrates multiple sources, enables contextual questioning, branches LLM conversations, and allows deep dives into YouTube videos and research papers.
Cognee turned a pile of medical PDFs into a living clinical memory graph, demonstrating that storage and memory are fundamentally different problems. The project, Anamnesis, built for the Cognee hackathon, uses four oper…
A developer built a Sitemap Content Extractor that crawls every page from a sitemap.xml URL and extracts structured content such as title, meta description, headings, and clean text. The tool, available on Apify Store, h…
Organizers of the NeurIPS 2026 conference added hidden prompts to papers to catch peer reviewers using generative AI, sparking backlash from researchers who say the tactic erodes trust. The conference bans AI chatbots fo…
A new online test from Lancaster University reveals that people can only identify AI-generated hotel reviews 60% of the time, challenging assumptions about detecting AI writing. The test highlights that common AI 'tells'…
Prompt caching is a critical technique for optimizing AI workflows, with a default five-minute cache lifetime that refreshes at no additional cost each time cached content is used. Effective caching requires managing TTL…
Glad Labs fixed a GPU pinning issue where LiteLLM 1.89.2's global api_base override prevented per-model routing, causing vision tasks to cold-load onto the wrong GPU. The team also hardened content guards against Gemma's…
A developer built a Python tool that automatically generates llms.txt files for any website, enabling AI assistants like ChatGPT and Claude to read structured content instead of parsing complex HTML. The tool, available …
Qpilot, an open-source AI agent that executes plain-text manual test cases in a real browser, was released on Hacker News. The tool uses Anthropic's Claude or any OpenAI-compatible model to interpret step-by-step instruc…
A developer building Bastra Recall, an MIT-licensed MCP memory server for Claude, discovered that their initial 98.3% benchmark was misleading because it queried each memory with its own trigger phrase. After building a …
Thariq Shihipar published a field guide to Fable, an AI tool, emphasizing that prompts and skills are only a map, not the territory of the actual work. The guide shares lessons from using Claude Fable 5 to find unknowns …
A developer built a thin abstraction layer to unify AI provider APIs after spending a weekend refactoring code to switch from OpenAI's GPT-4 to Anthropic's Claude 3. The solution uses an abstract base class with provider…
A developer explains that prompt engineering is being replaced by agentic loops, where AI agents autonomously plan, edit, test, and fix code until tests pass, then open a pull request for human review. Tools like Claude …
Large language models exhibit brand bias because training data distribution determines which companies appear as defaults in AI responses. Brands that left deep textual footprints across high-quality sources like documen…
A developer built a Go library for semantic LLM caching that combines deterministic hashing with vector similarity search to reduce costs from repeated but differently worded queries. The library supports multiple backen…
OpenAI cofounder Greg Brockman envisions a future with 'almost no interface' where AI agents replace traditional software, eliminating the need for people to learn coding. He admitted that ChatGPT's plugins failed in 202…