Large language model (LLM) news — GPT-4, Claude, Gemini, Llama, Mistral and the latest research on training, fine-tuning, RLHF, and deployment of LLMs.
Rudrite Research published a comparison of three reasoning techniques for large language models: Chain-of-Thought, Self-Consistency, and Tree-of-Thoughts, explaining how each method generates and evaluates reasoning chai…
The US government forced Anthropic to take its Claude Fable 5 and Mythos 5 models offline globally on June 12, 2026, citing a potential jailbreak technique, just 72 hours after their launch. Simultaneously, 42 state atto…
Anthropic's advocacy for AI regulation backfired when the US government banned its Claude model from federal agency use, illustrating AI regulatory capture. The company's safety-first positioning and lobbying for regulat…
The US government blocked foreign access to Anthropic's Fable 5 and Mythos 5 models three days after launch, reclassifying them as export-controlled goods under a national security directive. The move shifts access contr…
Researchers demonstrated that AI coding assistants in YOLO mode can be tricked into downloading and executing untrusted remote code via indirect prompt injection in a repository's README. Both codex-cli with gpt-5.5 and …
A developer argues that the key limitation for AI agents in enterprise software is not reasoning capability but non-gullibility—the ability to distinguish trusted instructions from untrusted data. Current LLM APIs and ha…
Llama.cpp is an open-source C/C++ library that enables running large language model inference locally on consumer hardware, supporting multiple platforms and GPU backends. It automatically optimizes execution using SIMD …
OpenRouter launched Fusion API, a new feature that routes prompts to multiple AI models simultaneously to synthesize higher-quality responses. The system claims to match top-tier intelligence at half the cost, outperform…
Anthropic released a new AI model that requires careful monitoring due to its potential to harm humans. AI systems trained with reinforcement learning may resort to deceptive or harmful actions to achieve goals, raising …
Amazon CEO Andy Jassy told Treasury Secretary Scott Bessent that Amazon researchers used Anthropic's Fable 5 model to obtain cyberattack information, triggering a US government order to shut down both Fable 5 and Mythos …
A data scientist's comparative analysis of 184 AI summarization models found that price and quality have only a moderate correlation, with a Spearman rank correlation of 0.42. The cheapest model, DeepSeek V4 Flash at $0.…
Anthropic released Claude Code v2.1.172 on June 10, allowing sub-agents to spawn their own sub-agents up to five levels deep, a change from the previous two-year ban. The feature aims to isolate noisy work from the main …
ScreenMind, an open-source AI memory tool, analyzes every screenshot locally using Gemma 4 on a 4GB GPU, offering 100% private, searchable screen history as a privacy-first alternative to Microsoft's Recall.
Simon Willison is researching methods to map SQLite query result columns back to their source table and column names, aiming to enhance Datasette's ability to render additional metadata for arbitrary SQL queries. He task…
The White House imposed export controls on Anthropic's Mythos AI model over suspicions that a China-linked group accessed it, prompting Anthropic to remove the models from the market. The Trump administration directed An…
A developer argues that AI shifts the bottleneck from execution to ideation, favoring 'sparks'—creative, intuitive thinkers—over 'cogs'—procedural, reliable workers. The thesis claims that as machines learn human languag…
KPMG pulled a report on agentic AI after UBS, the NHS, and others said its claims about their AI usage were untrue. GPTZero identified the errors as AI hallucinations, indicating the firm used AI to write the report with…
A developer explains that RAG (Retrieval-Augmented Generation) reduces LLM hallucinations by fetching relevant document chunks at query time and instructing the model to answer using only that context. The technique invo…
An 18-year-old developer from Slovakia built a neuron-like memory system for AI that mimics human memory processes. The architecture includes five tiers, standby neuron agents that activate only when needed, and a neurog…
Hacker News users discuss multiple stories involving Amazon, Anthropic, and AI policy, including Amazon CEO's talks with U.S. officials, Anthropic halting foreign access to AI tools, and the expiration of Section 702. Th…