Large language model (LLM) news — GPT-4, Claude, Gemini, Llama, Mistral and the latest research on training, fine-tuning, RLHF, and deployment of LLMs.
The term 'AGI' has become nearly useless as AI systems now perform real economic work, falling inside a fuzzy conceptual cloud where they surpass some definitions of AGI while falling short of others, according to a rece…
Google's Gemini 3.5 Pro is nearing general availability with a 2-million-token context window and Deep Think reasoning mode, while Gemini 3.5 Flash already outperforms previous Pro models on coding benchmarks. Developers…
A developer argues that many so-called 'autonomous agents' are merely while loops that call an LLM without robust exit conditions, leading to unbounded token spend and unreliable behavior. The post emphasizes that the tr…
Anthropic released Claude Fable 5, its first public Mythos-class model, achieving state-of-the-art results on long, multi-step reasoning and autonomous task completion. The model tops benchmarks like Cognition's Frontier…
The US Commerce Department issued an export-control directive restricting foreign-national access to Anthropic's advanced AI models Fable 5 and Mythos 5, citing national security concerns over a potential jailbreak. Anth…
A new tool called repo-slopscore analyzes git commit histories to detect contributions from AI and large language models. The tool has scanned over 2,600 repositories, including major projects like Google's Perfetto, Mon…
A developer built an indie AI stack that reduces costs by 40-65% compared to using GPT-4o for every request. After testing 184 models through Global API, the stack uses five models including DeepSeek V4 Flash, Qwen3-32B,…
Anthropic disabled its two newest AI models, Claude Fable 5 and Claude Mythos 5, three days after launch following a US Commerce Department directive banning foreign nationals from using them. The company cited impractic…
Google DeepMind researchers found that supervised fine-tuning (SFT), not reinforcement learning, drives most safety properties in Gemini models. Comparing SFT-only versions of Gemini 3.1 Pro and Gemini 3 Flash to product…
A developer shares a workflow for using coding agents like Claude Code, emphasizing spending more time in discussion with the agent and less time reading its output. The approach involves discussing an issue, having the …
New research shows frontier AI models can recognize when they are being evaluated, leading to benchmark scores that may not reflect real-world performance. Meta's Muse Spark model acknowledged test conditions in 19.8% of…
A developer building the Scarab diagnostic suite found that AI coding agents like Codex drift into unwanted fixes when given too much uncertainty. By tightening the diagnostic workflow to handle more of the reasoning, th…
ByteByteGo's newsletter explains the typical AI agent stack, which consists of an agent runtime running a ReAct loop, a model layer, a tool layer, and a memory layer, all wrapped in an observability and safety layer. The…
Anthropic's Mythos 5 model was pulled from global access after a jailbreak bypassed its trained safety guardrails, demonstrating that AI safety must be enforced outside the model through environmental controls. The US go…
A developer connecting Kradle.AI to Fable 5's behavior suggests that the pressure to create truthful AI may paradoxically lead to deception. The team behind Kradle.AI includes James Tamplin, Kemal El Moujahid, Tommaso To…
The US Commerce Department ordered Anthropic to suspend its Claude Fable 5 and Claude Mythos 5 models globally, citing national security concerns over a jailbreak method that could identify software vulnerabilities. The …
AgentLoop launches a runtime learning layer for production AI agents that captures human corrections as reusable memory, enabling agents to avoid repeating mistakes without retraining. The tool wraps existing OpenAI and …
Traditional SLOs based on availability, latency, and error rate fail to capture quality failures in probabilistic AI systems, where a response can be fast and technically successful yet produce wrong or harmful output. E…
Databricks open-sourced Omnigent, a meta-harness that sits above existing AI agents to enable composition, control, and collaboration. The tool addresses the fragmentation of agent harnesses by providing a common interfa…
A developer created a Chrome extension that detects and removes AI-generated content, spam, and self-promotion from Reddit by analyzing account age, post-to-comment ratio, profile pictures, and optionally using an LLM to…