Large language model (LLM) news — GPT-4, Claude, Gemini, Llama, Mistral and the latest research on training, fine-tuning, RLHF, and deployment of LLMs.
Estonia's Institute of the Estonian Language tested 60 AI models against Russian disinformation, finding all four versions of Mistral AI's models scored below 40%, with the best ranking 47th. The benchmark, released June…
NVIDIA released a developer example for building a transaction foundation model using accelerated computing, demonstrating a near-50% lift in fraud detection Average Precision over an XGBoost baseline. The model, pre-tra…
OpenAI reported $13.07 billion in revenue and $34 billion in costs for 2025, resulting in a $20.92 billion operating loss, as the AI giant prepares for an IPO expected in 2026. The company's expenses more than doubled fr…
The Economist published an article questioning whether AI wrote the piece itself, exploring the capabilities and limitations of AI-generated text. The article highlights the growing sophistication of large language model…
A developer abandoned Notion AI after its pricing ballooned, opting for open-source alternatives. Benchmarking showed Notion AI's optimized 2026 stack offered 40-65% cost reduction but relied on community models like Dee…
A developer built a bridge to connect ChatGPT's custom MCP connectors with the local web scraping tool CrawlForge, since ChatGPT requires remote servers. The solution uses a thin FastMCP wrapper that exposes CrawlForge's…
Andrej Karpathy stated in 2023 that language models do not know they are wrong, as they lack an internal uncertainty signal. A developer at Datawise argues that LLM errors are not hallucinations but rather the result of …
A developer debugging a production-only 500 error on their blog discovered that GitHub Copilot had introduced infinite recursion in a Markdown library. The bug only manifested on the production server, requiring creative…
JetBrains open-sourced Mellum2, a 12B Mixture-of-Experts coding model under Apache 2.0, designed for air-gapped and compliance-locked environments where external API calls are prohibited. The model uses a MoE architectur…
The US government forced Anthropic to disable access to its Mythos and Fable AI models after Amazon CEO Andy Jassy reported national security risks, citing concerns that China had accessed Mythos. Anthropic had refused t…
OpenAI will debut ChatGPT on the Pentagon's GenAI.mil platform in early July, making the chatbot available to over 3 million defense personnel. The company is working with the DoD's Chief Digital Artificial Intelligence …
A developer demonstrated that by pointing the OpenAI SDK at the base URL of api.airforce, a single client can access models from Claude, GPT, and Gemini by simply changing the model string. The gateway provides cross-pro…
A developer used an LLM to optimize a C collision detection implementation, achieving over 100x speedup from the baseline at a cost of ~125 million tokens. The optimization targeted Kevin Tracy's differentiable collision…
OpenAI has developed a method called Deployment Simulation that replays previous conversations with a new model to predict its behavior before release. In tests with GPT-5.4, the approach forecasted changes in production…
The U.S. government imposed export controls on Anthropic's Fable and Mythos AI models after Amazon researchers jailbroke Fable's cybersecurity guardrails, forcing Anthropic to disable the models for all users. Anthropic …
Mercury launched Mercury Command, a conversational AI interface for its financial platform, enabling users to perform tasks like checking cash positions and sending invoices via natural language. The feature requires exp…
European leaders will meet with U.S. President Donald Trump and Anthropic CEO Dario Amodei at the G7 summit in Evian-les-Bains, days after the U.S. suspended EU access to Anthropic's latest AI models. The move has intens…
AI may amplify the Dunning-Kruger effect, worsening societal overconfidence and misinformation, especially on specialist topics where AI regurgitates false claims from users.
A new pricing comparison reveals that premium AI chatbot plans from OpenAI, Google, Anthropic, and others range from $8 to $300 per month, offering varying levels of access, advanced reasoning, and agentic capabilities. …
Ruflo and MetaHarness have integrated an OIA-layered walkthrough across 28 iterations, implementing ADR-150 to improve AI-agent platform routing. The new learned router uses k-NN, kernel ridge regression, and FastGRNN to…