Large language model (LLM) news — GPT-4, Claude, Gemini, Llama, Mistral and the latest research on training, fine-tuning, RLHF, and deployment of LLMs.
GitHub Copilot experienced degraded availability for the Claude Opus 4.7 model across Copilot Chat, VS Code, and other Copilot products on June 8 due to an issue with an upstream model provider. The incident was resolved…
Amazon Web Services has released official Model Context Protocol (MCP) servers for Amazon Aurora MySQL, Aurora PostgreSQL, and Aurora DSQL, enabling natural language database queries and administration. The servers conve…
NVIDIA Jetson Orin Nano Super 8GB benchmarks show 25W as the energy-efficiency sweet spot for sub-4B Bonsai LLMs, delivering 47-48% more tokens per second than 15W while maintaining or improving output tok/J. The Ternary…
A new concept called "Judgment Compression" is emerging as a critical missing layer in production AI system design, addressing the gap between generating outputs and validating their structure. While validators can confi…
AI Gateway data shows DeepSeek's share of tokens on the platform surged from under 1% to 17% in May, while its share of spend remained near 1%, as the low-cost model entered production use. Anthropic's share of spend gre…
A developer built a continuous, durable personal AI agent called TorkBot after being inspired by Clawdbot. The agent maintains state across interruptions, allows mid-task steering, and resumes work after crashes, address…
A developer rebuilt Cloudflare's AI code review system as a 250-line JavaScript workflow using Claude Code, replacing Cloudflare's infrastructure with a single file that runs locally. The system uses specialized agents f…
Speculative decoding, a lossless inference optimisation that predicts future tokens to reduce latency, faces new economic constraints as modern mixture-of-experts (MoE) architectures replace dense transformers. MoE layer…
OpenRouter offers a single API endpoint for 341 models across 60+ providers with automatic failover, unified billing, and zero data retention without enterprise contracts, while direct APIs from Anthropic, OpenAI, and Go…
Thoughtbot engineers built a Claude Code hook that automatically runs RuboCop against any Ruby files an AI coding agent modifies, then gives the agent a chance to fix violations it can catch. The hook enforces team codin…
Spidra launched eight free developer tools for working with web data, including utilities for converting websites to Markdown or JSON, taking screenshots, and generating AI questions from page content. The tools require …
Developers building AI search or RAG pipelines often struggle to extract structured data from websites with inconsistent page layouts. A new approach uses a scraping layer that handles AI extraction natively, converting …
A developer building a basic AI agent from scratch adds long task planning capabilities by implementing an in-memory scratchpad and to-do list, enabling the agent to decompose complex tasks and track progress autonomousl…
A software developer explains why they refuse to use LLMs for coding, citing enjoyment of the craft, autonomy concerns, and the proliferation of low-quality AI-generated open-source contributions. They advocate for label…
A developer warns that AI-generated tests often contain 'rotten green tests' that pass without actually verifying anything, citing research showing up to 62% of LLM-generated tests fail to compile and that AI-co-authored…
Agentic search can be implemented in three distinct ways: retrieval-centric, harness-centric, and model-centric. The retrieval-centric approach relies on building high-quality search to guide agents, but search remains d…
Developer Joe Pohnelt created an email triage system using Claude Code subagents to automate Gmail management. The system uses six specialized AI agents for security analysis, relationship mapping, content summarization,…
Simon Willison released datasette-agent-edit 0.1a0, a base plugin for Datasette Agent that provides core text editing tools for agentic editing of files. The plugin implements view, str_replace, and insert tools modeled …
NAVER announced it will expand sovereign AI infrastructure using the NVIDIA DSX platform, starting with a 55-megawatt expansion at its GAK Sejong data center with plans to scale to gigawatt capacity. The South Korean AI …
A developer published a CLAUDE.md configuration file for Claude Code that enforces two working principles: judge work by correctness rather than ROI, and fix bugs at the architectural root rather than the symptom. The ru…