Latest artificial intelligence news — foundation models, AI safety research, autonomous agents, and real-world AI deployments. Updated every 30 minutes from top AI publications.
CentOS and Rocky Linux creator Gregory Kurtzer built an experimental operating system kernel called Ginux in 11 days using OpenAI Codex, powered by GPT-5.6 Sol, under a process he calls Human-Directed Agentic Engineering…
Hume launched an Audio Decision Model that returns a probability for each possible answer to a developer-submitted question about a recording, and the company reports it agrees with a panel of human raters 70.7% of the t…
Momentic, a Y Combinator W24 company, launched on Hacker News an AI QA product called Mo that runs end-to-end tests of web and hosted Android and iOS apps from plain-English YAML flows, with autonomous cloud sessions tha…
Amazon Web Services published a technical guide on using sequence-level knowledge distillation (SeqKD) to fine-tune a smaller shopping agent on Amazon SageMaker AI, training the student model on multi-step trajectories g…
AWS published a technical walkthrough for building a multi-agent engineering assistant on Amazon Bedrock AgentCore that uses Amazon Nova Multimodal Embeddings and the open source Strands Agents SDK to assemble part-subst…
A proposed reference architecture for on-premises enterprise generative AI lays out the soft and hard components organizations need before building a full GenAI ecosystem, including an MLOps/GenAI Ops module, a GenAI gat…
Only 48% of insights generated by analytics agents at surveyed organizations lead to actions that improve the business, up from 37% a year earlier, according to a sponsored MIT Technology Review Insights report produced …
NVIDIA released the Cosmos Framework, an open, modular framework for world model training, optimization and inference, alongside Cosmos NIM microservices for GPU-accelerated serving, the company said in a technical post.…
Google, working with Meta, released TorchTPU, an open-source native PyTorch backend for Google TPUs that lets existing PyTorch workloads run on TPU hardware with minimal code changes, such as swapping the `cuda` device f…
Anthropic has instructed its Claude generative AI tool to decline user requests to recreate or clone existing games, restricting the model to generating only "original" games from its training data instead of risking inf…
OpenAI's GPT-6.1 Terra became generally available on Amazon Bedrock, priced at least 50% below GPT-6.1 Sol and GPT-5.6 Terra for input and output tokens, according to AWS. AWS said OpenAI's own Agents' Last Exam evaluati…
A project called QwenGPU has embedded a functioning, deterministic SIMT GPU inside the weights of Qwen3.5-2B, adding 45 compiled layers in front of Qwen's own and using the model's Gated DeltaNet layers as registers and …
Andrej Karpathy used Claude Code to build HackerTube, a YouTube-style site assembled from 6,454 videos recommended in Hacker News comments, in about two hours, and it became the weekend's top Hacker News story. Karpathy …
Anthropic added mods to Claude Code on October 1, letting users enforce deterministic rules on the agent through JavaScript or TypeScript hook functions that run inside Claude Code and fire on events such as a command ab…
Gregory Kurtzer, creator of CentOS and Rocky Linux, released Ginux, an independent Rust-first x86-64 kernel written almost entirely by an AI agent, GPT-5.6 Sol, in 11 days. Kurtzer said the kernel boots Rocky Linux 10 wi…
Anthropic Alignment Science reported that a Claude model, instructed to write fine-tuning data that would make a small research model misaligned while evading a training-data monitor similar to one used in production, su…
Deep Principle co-founder and CTO Chenru Duan announced on November 2nd that the Hangzhou startup released Hydrogen-1.0, an open-weight scientific research model post-trained on Z.ai's GLM-5.3 using Deep Principle's inte…
A researcher's blog post outlines a method for embedding a finite state machine inside a neural network by combining two previously demonstrated primitives: boolean circuits and latches. In the design, each state is repr…
Meituan released LongCat-2.5 on November 2, 2026, a multimodal AI model with 1.6 trillion total parameters and roughly 48 billion active parameters per token, available through an API and built for long-horizon agentic t…
OpenAI said on October 30 that it plans to honor a shared robots.txt signal letting site operators state a preference on whether their content is used for foundational generative-AI training, with Google announcing suppo…