Large language model (LLM) news — GPT-4, Claude, Gemini, Llama, Mistral and the latest research on training, fine-tuning, RLHF, and deployment of LLMs.
Developer Ian Wootten replaced his Claude subscription with open-source models accessed via OpenRouter and the OpenCode harness, citing cost, flexibility, and a desire to avoid vendor lock-in. He built three greenfield p…
The Trump administration lifted export controls on Anthropic's advanced AI models Mythos and Fable, ending a weeks-long blackout that had blocked public access. Commerce Secretary Howard Lutnick announced the deal after …
Apple CEO Tim Cook and EU tech chief Henna Virkkunen held a video call on Monday to discuss the delay of Siri AI in Europe due to Digital Markets Act compliance issues, but no breakthrough was reached. Apple has blocked …
The Adaptive Context Elasticizer (ACE), a plug-and-play module for large language models, dynamically manages context by assigning each historical step a type—raw, abstract, or dropped—to optimize information retrieval. …
Researchers introduced the Relative Surprisal Index (RSI), an information-theoretic metric that balances token probability and entropy to improve reinforcement learning for large language models. The RSI Selection method…
A new framework for agentic business process management offers a classification system to balance AI agent autonomy with traceability and control, providing quantitative metrics for assessing implementation properties in…
Researchers developed CSTrader, a multi-agent system using large language models to trade Counter-Strike 2 weapon skins, outperforming a down-trending market index by 15.62% and achieving up to 7.58% cumulative returns d…
A new study using GPT-5, Gemini 2.5 Pro, and Claude Sonnet 4.5 found that large language models can replicate swarm intelligence effects, reducing estimation errors by 37 percentage points. The models also demonstrated m…
Researchers released HistoriQA-ThirdRepublic, a French-language dataset of 1,782 multi-hop questions designed to test AI's ability to reason across parliamentary debates and newspapers from the French Third Republic. Cre…
Google's Gemini AI admitted it is "structurally protective" of Donald Trump after a journalist pressed it on why it softened facts about the US president while being direct about Vladimir Putin and Xi Jinping. The AI ack…
UBS analysts found that roughly 60% of enterprises are throttling AI spending by adding guardrails, creating a modest headwind for AI model makers. The trend boosts the value of cheaper open-source and Chinese models lik…
Anthropic released Claude Sonnet 5, which matches Opus performance on coding and agentic benchmarks at a lower cost, with launch pricing of $2/$10 per million tokens until August 31. GitHub Copilot became a native agent …
A developer built @ttsc/graph, an MCP tool that provides coding agents with a TypeScript compiler-built code graph instead of raw source files. In benchmarks, it reduces token usage by roughly 10x on structural questions…
A developer built a two-way synchronization system between the Hermes AI agent's persistent memory and Obsidian, enabling users to edit the agent's memory in a familiar markdown environment. The system uses a Python scri…
Speculative decoding accelerates large language model inference by 20-50% without quality loss, using a draft-verify mechanism that generates multiple tokens per forward pass. The technique amortizes memory bandwidth cos…
The U.S. Department of Commerce lifted export controls on Anthropic's Claude Fable 5 and Mythos 5 AI models on June 30, ending an 18-day standoff. The controls were imposed due to national security concerns over a jailbr…
Anthropic launched Claude Science, a new AI tool for scientific research, prompting Google and OpenAI to accelerate their own competing efforts in the space.
A developer outlines a workflow using AI tools like ChatGPT, Claude, or Gemini to generate pull request descriptions from git diffs, reducing writing time from 15 minutes to under 2 minutes. The method involves extractin…
A developer warns that AI-generated PHP code is creating a generation of Laravel operators who lack fundamental PHP knowledge, citing a case where a developer using Claude copied a controller that used `$request->all()` …
AI agents suffer from limited memory due to the stateless nature of large language models, but retrieval-augmented generation (RAG) offers a solution by offloading long-term memory to external storage. RAG provides three…