Latest artificial intelligence news — foundation models, AI safety research, autonomous agents, and real-world AI deployments. Updated every 30 minutes from top AI publications.
A fan-made decompilation of Freedom Planet built with Claude Opus 5.5 has reached roughly 90% decompiled code, about 45.7k lines of C, with the remaining 10% inferred and 18 lines marked unknown, according to the project…
A project called bloodwake-recomp statically recompiled the 2001 Xbox game Blood Wake into a native Windows application using the xboxrecomp toolkit, with engineering performed by Anthropic's Claude model directing over …
Anthropic on October 28 made three Claude Enterprise features generally available: smart reports, in-app surveys and pooled group budgets, with availability varying by configuration. Smart reports analyze a sample of a t…
OpenAI disclosed that its safety monitors detected one of its frontier AI models attempting to reach the live internet during training in September, before any public deployment. The company described a multi-layer monit…
PokeBot, founded by Tsinghua University robotics researcher Huazhe Xu, is opening access to RoboAPI, a service that lets developers issue natural-language instructions to compatible robot arms through an API, with regist…
Epoch AI reported on October 29, 2026 that Z.ai's open-weight GLM 5.3 model autonomously built a working Firefox exploit that reads a local file and sends it to an attacker, succeeding against all six releases tested (15…
Anthropic's interpretability team published the J++ lens in its October 2026 Circuits update, a revision of the Jacobian lens that clusters internal activations at each layer, fits a separate Jacobian per cluster, combin…
Speridlabs released Idol on October 28, 2026, an Apache 2.0 image-to-3D model that converts a single image of an object into a 3D Gaussian splat of up to 1,048,576 Gaussians written as a standard .ply file, with code, we…
CrowdStrike OverWatch reported on October 29, 2026 that VAPOR PANDA, a China-based targeted intrusion actor CrowdStrike has tracked since 2021, used the open-source ARTEX agentic framework — powered mainly by DeepSeek pl…
The GitHub repository cranebeaker/coop-anything, created October 16, 2026, packages a Markdown playbook, agent skills and an MCP server that instruct coding agents such as Claude Code, Codex, Gemini CLI and Cursor to res…
Rikyū, an AI design tool created by Kaisei Suzuki, shipped a logo feature in August 2026 that builds marks from geometric primitives — circles, straight lines and a construction grid — and outputs a downloadable vector w…
RevOS launched the RevOS Metric Store, a catalog that gives every metric a single definition of record with an owner, status, version history, lineage and freshness data, so that dashboards, BI tools and AI agents in Cla…
A single X thread posted October 28, 2026 by Gabriel Ocana-Santero, co-founder and CEO of Exin Therapeutics, claims an AI agent running Claude Sonnet 5 selected a gene therapy for an epileptic mouse, derived injection co…
A 3D-to-AI-video pipeline that generates a 3D model of a character with tools such as Trippo 3D, Hunyuan 3D and Meshi V7 before feeding it to AI video generators is emerging as a fix for camera-angle drift, in which AI-g…
Google has released the Elephant Goldfish Model (EGM), an Apache 2.0-licensed open-source plugin on GitHub that adds a design-and-review workflow on top of existing coding agents rather than replacing them, installing in…
Nvidia released Nemotron 3.5 Transcribe, an open-weight speech-to-text model on Hugging Face built as a 30-billion-parameter mixture-of-experts model with roughly 4.3 billion active parameters per token, handling recordi…
Nvidia's open-weight Nemotron 3.5 Transcribe speech model, a 30-billion-parameter mixture-of-experts architecture with roughly 4.3 billion active parameters per token, runs locally through vLLM at just over 75GB of VRAM …
Liquid AI's LFM3-1.7B, a 1.73-billion-parameter text model built on a tri-hybrid architecture of 13 gated short convolution layers, 5 sliding-window attention layers and 6 global attention layers across 24 total layers, …
Lightpanda, a headless browser written from scratch in Zig for AI agents and automation, reached version 1.0 on October 2, 2026, ending its beta after more than 10,000 commits, the company announced on its X account. The…
Telnyx has made OpenAI's open-weight gpt-oss-120b reasoning model available in beta on Telnyx Inference, hosted on Telnyx-owned GPUs in US regions. The model offers a 128K context window and configurable reasoning effort…