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 tested LLM code translation by passing a Go program through 10 languages and back, finding the final Go version grew from 94 to 443 lines while retaining correctness. The experiment used a chain of LLM-genera…
A developer argues that production AI agents require a runtime layer, not just a framework, to handle durability, isolation, resource control, and lifecycle management. Without a runtime, agents fail in production due to…
OpenAI has secured Samsung Electronics as a major customer for ChatGPT Enterprise and Codex, making the AI tools available to all employees in Korea and the Device eXperience Division worldwide. The deal expands OpenAI's…
A classic psychology test has revealed that large language models, including GPT-5, Claude, and Gemini, struggle to maintain focus when faced with competing information over extended periods. Researchers led by Suketu Pa…
Jaya Baloo, COO & CISO at Aisle, argues that restricting access to cyber-capable AI models misunderstands how attackers and defenders operate, and that open-weight models benefit both sides. She warns that limiting acces…
AI red teaming must target runtime boundaries, not just prompt tricks, to prevent failures like prompt injection, confused deputy attacks, and memory contamination. A disciplined loop of probing and architectural fixes i…
Tencent is testing an AI agent inside WeChat that can complete tasks across mini-programs, leveraging the super app's 1.4 billion users and existing service ecosystem. The agent aims to integrate AI into daily habits rat…
A developer tracked down a 36GB memory leak in a Claude Code memory server caused by sql.js WebAssembly filesystem (MEMFS) not being cleaned up. The leak occurred because garbage-collecting the JavaScript Database wrappe…
A developer built OpenHop, a new output format for AI coding agents that allows step-by-step playback of system explanations, addressing the problem that agents generate explanations faster than humans can verify them. T…
A developer explains that in production AI systems, the user's actual prompt often accounts for less than 5% of the total context sent to the model, with the rest comprising system instructions, retrieved documents, conv…
Researchers from Wake Forest University found that 282 of 444 iOS apps with large language model features leaked AI credentials or exposed backend access mechanisms, with only 28% fixing the issues after a 90-day disclos…
A developer built a personal intelligence system that curates GitHub and news using an LLM, delivering three scheduled briefings to Telegram. The system, hosted on a budget VPS, uses OpenRouter for LLM calls and includes…
MemoryOps AI launched an open-source governed memory infrastructure for AI assistants, implementing a ChatGPT-style memory lifecycle with policy enforcement, typed storage, hybrid retrieval, and auditability. The system …
Sakana AI's Fugu is a multi-agent orchestration system that coordinates multiple frontier language models through a single API, dynamically routing tasks to specialist agents. The system, grounded in ICLR 2026 papers TRI…
CSGHub, an open-source platform for managing Large Language Model (LLM) assets, offers on-premise deployment for secure, offline operation. It provides a unified interface for uploading, downloading, storing, and distrib…
Moonshot AI released Kimi K2.7 Code, a 1-trillion-parameter open-source coding model that uses 30% fewer reasoning tokens than its predecessor and outperforms Claude Opus 4.8 on MCP tool-calling benchmarks. The model is …
A developer released Recall, a free offline plugin for Claude Code that provides persistent session memory without API calls or cloud dependencies. The plugin uses a local append-only log and a classical Python summarize…
Developer kromych released badc, a cross-platform optimizing C compiler built with Claude that supports C99, C11, and later standards. The compiler can build Python 3.14 on five targets, JIT-compile code, and produce nat…
MiniMax introduced sparse key-value caches that reduce attention scaling from quadratic to near-linear, enabling practical multi-hundred-kilobyte context windows on a single GPU. The method cuts per-token attention compu…
Vercel open-sourced Eve, a framework for building durable AI agents with a filesystem-first approach. A developer tested it by building a shopping assistant that can search catalogs, check inventory, compare prices, read…