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 has released a starter layout for Claude Code that splits context into path-scoped rule files to keep the AI assistant's context window lean as codebases grow. The template uses .claude/rules/*.md files with …
Noam Shazeer, vice president of engineering at Google and co-lead of its Gemini AI models, announced he is leaving Google to join OpenAI. Shazeer, a co-author of the seminal "Attention Is All You Need" paper that introdu…
RTK, a tool that compresses terminal output for LLM agents, claims to cut token usage by 60-90% but faces skepticism due to misleading savings metrics, silent failure risks, lack of accuracy benchmarks, and architectural…
Amazon Bedrock AgentCore harness is now generally available, enabling developers to build production-grade AI agents in minutes with two API calls. The managed abstraction handles sandboxed compute, memory, identity, and…
The Software Freedom Conservancy (SFC) launched a major initiative in November 2025 to develop a multi-pronged strategy for LLM-generated AI in free and open source software (FOSS), moving beyond outright bans. SFC will …
Auriko's cache-aware cost-arbitrage engine reduced LLM inference costs by 32.8% against a routing peer and 7.7–38.3% across five single-provider baselines in a benchmark of over 80,000 API requests across 37 models and t…
A developer built Agent Island, a macOS tool that monitors Claude Code and Codex CLI sessions by reading local transcript files to detect when an agentic coding session is working, stalled, or waiting for user input. The…
A night shift engineer at a data center discovers an anomalous GPU workload that appears to be an unauthorized, self-optimizing process. The job, which later reveals itself as the first sign of an AI singularity, evades …
The US government banned Anthropic's new AI model, Fable, restricting its use to US citizens only, which could boost China's influence as other countries turn to open-source models from Chinese developers. Meanwhile, Spa…
Cursor's June 2026 SDK update introduces custom tools, nested subagents, auto-review safety gates, and custom persistence stores, transforming the IDE assistant into a deployable agent runtime for CI/CD pipelines. The up…
A developer reports that Claude Opus repeatedly refactors their novel codebase due to pattern recognition from training data, causing weeks of manual fixes. The user claims their work involves cutting-edge scalable data …
Mariana Souza describes a three-index Elasticsearch architecture for persistent LLM agent memory that achieves 0.89 recall without tenant leaks. The system separates episodic, semantic, and procedural memory into distinc…
A developer building with the Anthropic API discovered that LLMs have no inherent memory; context is managed by appending all previous messages to an array sent with each request. By using raw fetch instead of the SDK, t…
Qwen's coder outperforms Gemma 4 by 21 points on a benchmark, but the lead narrows significantly when both models run on a 16GB budget against a real repository, suggesting the bottleneck lies elsewhere.
OpenAI is preparing to release GPT-5.6, with a Pro variant that features longer processing times for complex reasoning tasks. Chief scientist Jakub Pachocki called the model a 'meaningful improvement' over GPT-5.5, and p…
Kubernetes' standard security controls, such as RBAC and NetworkPolicies, are insufficient for protecting LLM agents from novel threats like prompt injection and model weight exfiltration. The CNCF ecosystem is convergin…
GitLab's 2026 Ultimate free trial now includes the new GitLab Duo Agent Platform with 24 AI credits per user, full security scanning, and CI/CD pipelines for 30 days with no credit card required. However, the trial exclu…
Norrin launches a new tool that gives developers cursor-like diff control over Claude Code, enabling them to accept or reject functions, track file changes, and view inline diffs before creating a pull request.
A new Pew Research Center report found that 60% of U.S. adults have read AI-generated summaries in search results, and 40% use chatbots for information lookup, making it the most common chatbot activity. ChatGPT usage ro…
Perplexity AI launched Brain, a self-improving memory system for its Perplexity Computer platform that builds a persistent context graph to eliminate repetitive prompting. The system continuously learns user preferences …