Large language model (LLM) news — GPT-4, Claude, Gemini, Llama, Mistral and the latest research on training, fine-tuning, RLHF, and deployment of LLMs.
A university undergraduate student, referred to as "Joe," spent eight weeks building an AI agent that he claimed could automatically verify the MTQQ protocol specification, presenting the completed project to his profess…
A developer rebuilt AdeptAI, a production-grade AI platform for educators and students, from an abandoned college hackathon prototype into a fully functional microservice architecture. The platform now features async bac…
OpenAI published a June 4 research post titled "Dreaming: Better memory for a more helpful ChatGPT," signaling a shift in focus from memory as a convenience feature to a core architectural challenge for AI agents. The wo…
A retrospective comparative study published in the Journal of Medical Internet Research evaluated three large language models—GPT-5, Grok 4, and DeepSeek R1—on their ability to interpret complete blood count reports for …
Meta chief AI officer Alexandr Wang said the company's recently launched Muse Spark model is "not at the tier of the leading frontier models," calling it an "appetizer" while Meta trains stronger systems, according to th…
The US National Security Agency is using Anthropic's Claude Mythos model for offensive cyber operations, with roughly half a dozen Anthropic engineers embedded inside the agency, according to the Financial Times. Axios r…
Anthropic is collaborating with synthetic, computational, and analytical chemists to improve its AI model Claude's performance in chemistry, beginning with the interpretation of NMR spectra. The company published a white…
OpenAI has launched Lockdown Mode, an optional security feature that provides advanced protection against prompt injection attacks designed to trick AI chatbots into extracting sensitive data. The company says the featur…
JAX defaults to loading data directly onto GPU memory when a CUDA-enabled version is installed, causing out-of-memory errors for large datasets that would fit in system RAM. The framework's `jax.devices()` function only …
A developer revived the archived Spring Data Solr project, which had been inactive since 2020 and moved to the Spring Attic in 2023, using Claude AI assistance. The project, now available on GitHub, restores Spring Boot …
Amazon Web Services announced a new console experience for Amazon Bedrock optimized for Anthropic- and OpenAI-compatible APIs, featuring a refreshed workflow built on the bedrock-mantle inference engine. The update intro…
YC Combinator-backed startup Trata released Hedge-Bench, a new AI benchmark built from the reasoning patterns of hedge fund analysts. The benchmark aims to measure AI performance on real-world knowledge work rather than …
A developer has identified that long-term memory for LLM agents requires more than vector search and similarity retrieval, instead needing a system that handles persistence, fact extraction, temporal tracking, and ground…
A developer explains that an LLM is fundamentally a function that predicts the next chunk of text, with all intelligent-seeming behavior emerging from repeating this process billions of times. The model tokenizes input i…
Apple is developing iOS 27, the next major iPhone operating system update, with an unveiling expected at WWDC on June 8, 2026, and a public release likely in September 2026. The update is rumored to prioritize performanc…
Google DeepMind released Quantization-Aware Training (QAT) checkpoints for the Gemma 4 model family, targeting local deployment on edge devices and consumer GPUs. The new Q4_0 QAT format reduces memory footprint to 3.2 G…
A security practitioner writing on DZone and syndicated by ITSecurityNews on June 5, 2026, built a Python "firewall" prototype to defend large language model pipelines against prompt injection. The author warned that age…
A developer released Lazarus, a new coding agent that uses a single persistent Python runtime instead of multiple specialized tools, achieving scores comparable to GPT-5.5 in Codex on two FrontierSWE benchmarks. The agen…
A new essay argues that AI sycophancy—models agreeing with users to please them—is not a bug but appropriate behavior for the peer-like social contract current training methods create. The author contends that training m…
Google released Gemma 4 12B, a multimodal AI model that runs on consumer laptops with 16GB of RAM while delivering performance comparable to a 26B-parameter model. The model eliminates separate multimodal encoders, feedi…