Large language model (LLM) news — GPT-4, Claude, Gemini, Llama, Mistral and the latest research on training, fine-tuning, RLHF, and deployment of LLMs.
Amazon AI chief Peter DeSantis admitted the company's AI models lag behind OpenAI and Anthropic, but expressed hope of catching up within a year. Amazon is investing $33 billion in Anthropic while building its own rival …
Amazon Senior Vice President Peter DeSantis acknowledged the company's AI models have not been at the frontier of the largest workloads but expressed hope that Amazon will be in the conversation about leading models with…
A developer outlines four memory architecture options for AI agents that lose context after multiple turns. The choices include in-context window, vector memory store, episodic memory with summarization, and Redis sessio…
Companies are manipulating AI chatbots to smear competitors by publishing biased reviews and listicles that get cited by AI answer engines. A report by Hard Numbers found that 70% of buyer-style prompts were influenced b…
By 2026, AI workloads will consume roughly 40% of enterprise Kubernetes clusters, but the default scheduler is ill-suited for GPU-intensive tasks, leading to 30-45% GPU utilization rates and wasted costs. Platform teams …
Social media platforms including Threads, Instagram, and TikTok are introducing AI-powered tools that let users personalize their recommendation algorithms, moving away from one-size-fits-all feeds. Threads launched a pr…
A Hacker News user asks how much of today's internet content is generated by large language models, noting a trend of AI creating content for other AI systems and questioning when the internet will become a closed loop.
Alberto argues that humans can remain irreplaceable in the AI era by focusing on skills that cannot be measured or described, such as creativity and lived experience. He notes that AI excels at tasks like math and coding…
TokenCap released v0.7 — Smart Retrieval Engine, which allows developers to query their codebase with natural language instead of manually assembling context for AI tools. The engine searches across project intelligence,…
A freelance developer built an AI email assistant for a SaaS client and discovered that using GPT-4o for email classification at production volume cost $200-$375 per month for 50,000 emails. By switching to Global API, w…
A developer building a Model Context Protocol (MCP) server for a course catalog learned to distinguish between Tools, Resources, and Prompts. The team refactored their V2 implementation, which incorrectly exposed read op…
Cursor launched Origin, an AI-native GitHub alternative designed for teams using multiple AI agents to generate code. The platform is Git-compatible but built to handle parallel changes from dozens or hundreds of AI agen…
A developer built a Claude Code plugin that routes lightweight AI tasks to specialized small language models (SLMs) to reduce compute costs. The plugin intercepts requests and directs simpler queries to smaller, cheaper …
AI agent systems face a memory crisis due to the lack of a coherent memory model, with context windows acting as flat caches that waste tokens on irrelevant data. Engineers must adopt memory hierarchy concepts from compu…
Z.ai released GLM-5.2, a 753-billion parameter open-weights large language model optimized for long-horizon coding tasks. The model features an IndexShare architecture that reduces per-token compute FLOPs by 2.9 times at…
BrainGrid, an AI-powered platform for developers, reduced API costs by implementing Anthropic's ephemeral tool caching in AI SDK v5. By adding three lines of code to mark tool definitions as cacheable, the company cut to…
Consulting firm KPMG withdrew a report on Agentic AI after AI detection company GPTZero found it contained AI-generated fabrications and mangled citations. The report, titled 'Total Experience: Redefining Excellence in t…
OpenAI's GPT-5.4, launched on March 5, 2026, is being used to optimize Chan-Lam coupling reactions in pharmaceutical synthesis, improving yields through AI-generated reaction parameters tested by human chemists. The mode…
Shrijith Venkatramana, a developer building git-lrc, explains that NVIDIA Collective Communications Library (NCCL) is the critical infrastructure enabling multi-GPU training of large language models. NCCL provides optimi…
TextStack, a .NET-based AI product, has implemented a two-part evaluation system: an offline gate that runs golden dataset evals as opt-in dotnet tests to catch regressions before shipping, and production monitoring that…