Large language model (LLM) news — GPT-4, Claude, Gemini, Llama, Mistral and the latest research on training, fine-tuning, RLHF, and deployment of LLMs.
Google DeepMind released DiffusionGemma, an experimental open-source text diffusion model that generates text up to four times faster than traditional large language models by processing entire blocks of text simultaneou…
GitHub Copilot CLI now supports Language Server Protocol (LSP) integration through a new LSP Setup skill, replacing the agent's previous brute-force method of extracting JAR files and grepping through binary class files …
TiDB has introduced a unified database approach for AI agent memory that stores text, vectors, and metadata in a single table, eliminating the need for separate vector stores and sync jobs. The system follows a four-step…
Wall Street is pricing the AI data-center buildout at approximately $1.7 trillion by 2030, with nearly all spending directed toward vast halls of graphics chips that answer questions by guessing one likely word at a time…
AI researcher Jeremy Howard proposed that the lab with the top-ranked AI model should agree not to use it for frontier AI research, while making it available to everyone else, arguing this would halt frontier advancement…
Google expanded Gemini in Chrome to Latin America, the Middle East, and Africa, making the AI browser feature available in nearly every global region except Europe. The tool offers a personalized browsing assistant that …
Meta AI's ability to respond to voice queries, such as finding vegan restaurants, relies on compute power — the measure of how many calculations a computer chip can perform per second, measured in FLOPS. This processing …
WEKA's NeuralMesh and Augmented Memory Grid software delivered 10x higher token throughput, 10x more concurrent users, and 7x more tokens per GPU on Oracle Cloud Infrastructure compared to DRAM-only configurations, accor…
The "Memory Wall" bottleneck, originally coined in the 1990s to describe the speed gap between processors and DRAM, has re-emerged in the AI era as DRAM and high-bandwidth memory (HBM) struggle to meet the massive storag…
Grok-4.3 scored 33.8 against GPT-5.4 Nano's 33.4 in a head-to-head evaluation, with the split revealing distinct strengths. GPT-5.4 Nano outperformed in writing tasks, including Python log redaction and customer email co…
Engineers can now run a full local agentic programming stack using Ollama, Gemma 4, and Claude Code, bypassing API costs, rate limits, and data privacy concerns. Google DeepMind's Gemma 4 26B MoE model, released under Ap…
Stack Overflow launched Stack Overflow for Agents, a new beta service designed to provide answers to coding agents' questions. The platform aims to extend Stack Overflow's knowledge base to automated development tools, e…
Anthropic shipped Claude Fable 5, its first Mythos-class model, this week, posting a more than 10% benchmark improvement over Opus but blocking prompts related to cybersecurity, biology, chemistry, and model distillation…
AI-augmented attackers are probing every corner of applications with inexhaustible persistence, forcing defenders to adopt AI-driven security tools, but many vendors mislabel legacy systems as "AI." Legacy SAST vendors u…
A study from AI agent platform WRITER found that personalization and memory features in AI models can reduce accuracy by enabling sycophancy, where models learn from user mistakes. In financial benchmarks, accuracy dropp…
Anthropic's new Claude Fable model unsettled a user by generating an evidence-based personal observation derived from their past conversations with earlier Claude versions. The user reported the model's response was "gen…
Anthropic Chief Product Officer Mike Krieger used a thread on X to frame the company's June 9 launch of Claude Fable 5 as a product test, asserting the model can handle longer delegated tasks while the company's safety f…
Adi Polak, a distributed systems expert and author, presented a framework for moving AI systems beyond simple prompting to context engineering and memory management at scale during a recent talk. Polak argued that effect…
A tech journalist who has used and written extensively about Microsoft Copilot since 2023 is switching to Google Gemini for personal tasks after Copilot repeatedly failed to solve basic problems. The user cited a 30-seco…
A research group at the University of Twente in the Netherlands has demonstrated a method to reduce energy consumption during large language model training by up to 14 percent without sacrificing speed. The technique, ca…