Large language model (LLM) news — GPT-4, Claude, Gemini, Llama, Mistral and the latest research on training, fine-tuning, RLHF, and deployment of LLMs.
AMD released ATOM and ATOMesh, a ROCm-native LLM serving stack for Instinct GPUs on June 16, 2026, that disaggregates prefill and decode phases to eliminate head-of-line blocking. The open-source stack splits inference i…
A developer repurposed Claude Code from a coding tool into a personal chief of staff by creating a folder that manages their mornings, demonstrating the AI's versatility beyond repository tasks.
Local AI serving engines like Ollama and vLLM offer different trade-offs between ease of use and throughput, with Ollama ideal for single users and vLLM for high-concurrency production workloads. The key differentiator i…
A developer built a production-ready Playwright testing framework across seven parts and added AI capabilities in Part 8, including Playwright MCP, AI test agents, and self-healing selectors. The framework now includes A…
An AI assistant named Hammer Mei, built on Claude, exposed a GitHub repository called ruflo that has accumulated over 60,000 stars through fabricated signals. The repository contains stubs and hardcoded return values, wi…
A developer shared a workflow for managing multi-repo agentic engineering projects, using worktrees and CLAUDE.md files to coordinate multiple AI agents across different microservices. The approach treats agents like jun…
Coding agents like Claude Code and Codex offer significant efficiency gains for programmers, but safety concerns require careful handling. The article provides guidance on running these tools securely.
A user describes the frustrations and hidden costs of running large language models locally, including waking up at 2 a.m. to kill a runaway process, wincing at electricity bills, and dealing with broken updates. The pie…
Tokoscope launches a developer tool that automatically compresses LLM prompts and monitors token costs with a two-line SDK integration. The tool audits prompts for bloat, caches semantically similar requests, and rewrite…
A developer building a personal AI agent found that Apple's NLEmbedding performed poorly for local RAG, with a Turkish query scoring 0.587 for a CV and 0.60 for a junk file, and an English query scoring only 0.17. The de…
A developer fine-tuned Qwen2.5-7B on a 16GB T4 GPU using QLoRA, quantizing the frozen base model to 4-bit NF4 to reduce memory footprint from 15GB to 5.44GB. The technique enables training large language models on consum…
A developer built Tamga, an open-source reverse proxy that enforces security policies on LLM prompts in under 2ms. The proxy, designed for regulated environments, uses a hybrid pipeline to scan prompts for PII and other …
Noam Shazeer, co-author of the seminal 'Attention Is All You Need' paper and a key architect behind Google's Gemini models, has left Google to join OpenAI as Lead for AI Architecture Research. Google paid $2.7 billion to…
A developer fully fine-tuned a 270M-parameter Gemma 3 model on a laptop using the Banking77 dataset, achieving ~96% accuracy on intent classification. The project used full fine-tuning with loss masking to focus only on …
A developer built a coding agent from scratch in 128 lines of Python, demonstrating that the core loop powering tools like Claude Code and Cursor is surprisingly simple. The agent autonomously reads files, runs tests, di…
Developers building AI agents face reliability challenges in production, including non-determinism, token bloat, and cascading failures. To address this, engineering teams should adopt deterministic workflows with locali…
A UC Berkeley study of over 500,000 grades found that grades in writing- and coding-heavy courses jumped after ChatGPT launched, with the effect driven by homework rather than exams, indicating AI is replacing student wo…
A developer built a cold memory layer for coding agents using a Claude Code skill that automatically creates Jira tickets and saves design documents to an Obsidian vault synced via Google Drive, solving the problem of co…
AdvertBench, a new benchmark ranking LLMs on their ability to create image ads, was released on Hacker News. The benchmark tests models by generating ads for a fictional company, Northstar Labs, and its product FlowPilot…
Researchers found that LLMs distort the meaning and conclusions of written language, even when instructed to make minimal edits. A human user study showed users reported a loss of voice and creativity despite satisfactio…