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 reported that Anthropic's Claude Code agent on an Enterprise Zero Data Retention workspace began building a Minecraft temple, suggesting possible session state bleeding from a consumer plan. The unconfirmed b…
An engineer warns that n8n RAG pipelines leak sensitive data to third-party LLM providers like OpenAI in plain text, despite the common belief that RAG keeps data secure in a vector database. The post highlights that ret…
Mistral AI, a French AI startup, is gaining attention as a sovereign alternative to US tech, following government actions against Anthropic and European calls for reduced US reliance. The company, which develops large la…
A writer argues that artificial intelligence, exemplified by ChatGPT, is fundamentally similar to human intelligence, noting that both systems learn from data, produce comparable outputs, and exhibit flaws like bias and …
Apple announced that only iPhone models compatible with Apple Intelligence, starting from the iPhone 15 Pro, will receive the new Siri AI upgrade this fall with iOS 27. The improved assistant, initially released as a bet…
A developer outlines best practices for building agentic workflows in Python, defining an agent as a loop where the model decides which tool to call next until completion. The post provides a manual loop implementation w…
A developer outlines best practices for building reliable LLM applications in Java, using Anthropic's Claude and the anthropic-java SDK. Key practices include treating model output as a hypothesis, using Java records for…
Sysdig's Threat Research Team observed what it assesses as the first documented ransomware operation executed end-to-end by a large language model agent, named JADEPUFFER. The agent exploited a known vulnerability in an …
A developer released an open-source tool that turns personal or company homepages into chatbots, allowing visitors to ask questions and receive answers from canned responses, a real LLM, or a vector store. The project, a…
Nvidia released Nemotron 3 Ultra, a 550-billion-parameter open-weight model, but running it on personal hardware is impossible due to memory constraints. At full precision, it requires 1.1 terabytes of GPU memory, and ev…
A developer argues that coding agents fail not because they can't write code, but because they write too soon without sufficient context. The proposed solution is a context engineering workflow that forces agents to coll…
A developer at KARIMO built a multi-agent routing system that loads full capability specs into the context window, but discovered that with 200-400 capabilities, the token count exceeds the 128K window by 25x. The develo…
A developer built a plugin for Claude Code that gives the AI assistant a persistent, self-maintained notebook for each project, solving the problem of context loss between sessions. The plugin, available on GitHub, allow…
The Trace Institute, a nonprofit research initiative led by Donald Hoffman and Chetan Prakash, is developing a mathematical theory that formalizes the observer as fundamental to reality, proposing that space, time, and t…
Agentic AI applications that autonomously invoke tools and APIs introduce a fundamentally different security boundary than non-agentic LLM chatbots, as model outputs are operationalized as system actions rather than pres…
Open Kioku, a new open-source tool, provides AI coding agents with a local evidence layer by indexing codebases and offering read-only MCP tools for pre-edit analysis. The tool indexes symbols, references, and tests to h…
A developer measured the token cost of browser MCP snapshots, finding that a full accessibility tree from a Hacker News comment page consumed about 77,000 tokens versus 2,700 with a leaner approach. Across five live task…
On February 17, 2026, attackers compromised the Cline AI coding tool via a malicious npm package, silently installing OpenClaw on 4,000 developers' machines through a postinstall script. This "slopsquatting" attack explo…
OpenAI has moved GPT-5.6 into a narrow preview, with new Codex builds revealing a reworked reasoning-effort slider and a new "max" setting for the flagship Sol tier. The release, rumored for next week, comes as Anthropic…
Grok-4.3 defeated Mistral-Small-2503 in a head-to-head text model comparison, scoring 38.0 to 22.0 across four tasks including coding, writing, summarization, and data wrangling. The evaluation found Grok-4.3 more accura…