AI news — 15 November 2026

Every morning we collect the most important AI news from dozens of sources and sum it up in a few lines, with the newest open-source projects beside it. Each card links to the original source.

RSS · Today's newsRSS · New repositories

News 8 stories

The Decoder
Models & releases

Google's Gemini 4 Carbon Tops MathArena Benchmark Ahead of OpenAI

Google's Gemini 4 Carbon has taken the top spot on the MathArena math benchmark ahead of OpenAI's GPT-6.1 Sol, while Anthropic's Claude Fable 5.2 placed second. Although Gemini rarely gave up on tasks, its most common failure in testing was exceeding the time limit. This performance comes with a much higher price tag, costing roughly 15 times more than GPT-6.1 Sol under the tested settings.

Original

Gemini 4 Carbon takes the MathArena crown but costs 15 times more than GPT-6.1 Sol

For the first time in a while, no OpenAI model leads the MathArena AI math benchmark. Google's Gemini 4 Carbon and Anthropic's Fable 5.2 have knocked GPT-6.1 Sol from the top, though both cost significantly more. Gemini ignored the time limit in testing and almost never gave up on a task. The article Gemini 4 Carbon takes the MathArena crown but costs 15 times more than GPT-6.1 Sol appeared…

The Decoder
Research

OpenAI Reorganizes Robotics Team as General-Purpose Models Show Promise

OpenAI has reassigned its robotics researchers to its primary research and applied divisions rather than maintaining a dedicated unit. The shift comes as the company finds general-purpose models like GPT-6 Astra capable of directly controlling robots without specialized AI. Recent simulation benchmarks demonstrated that Astra outperformed dedicated robotics models across multiple tasks.

Original

OpenAI reassigns robotics researchers to other teams as general-purpose models might make specialized robotics AI unnecessary

OpenAI is dissolving its standalone robotics team for the second time, but not because it lacks data. General-purpose models like GPT-6 Astra can apparently control robots directly, making a dedicated unit unnecessary. The researchers are being redistributed across other teams in what the company frames as a way to accelerate its robotics efforts. The article OpenAI reassigns robotics…

The Decoder
Models & releases

Runway open-sources Praxis-1 AI model for robot control

Runway has released Praxis-1, an open-source robotics model under the Apache 2.0 license that uses knowledge learned from video training to direct robot movements. By taking camera feeds, text instructions, and position data, the model predicts environmental changes and continuously adjusts controls. Although it leads its category on Nvidia's RoboLab-120 benchmark, it still requires specialized training footage and cannot independently plan complex action sequences.

Original

Runway's open-source Praxis-1 turns video AI into a robot controller

Runway releases its robotics model Praxis-1 with open weights and source code under the Apache 2.0 license. The model uses knowledge from video training for robot control and leads the World Action Model category on Nvidia's RoboLab-120 benchmark. It can't yet plan complex action sequences on its own, though. The article Runway's open-source Praxis-1 turns video AI into a robot controller…

TechCrunch
Business

Jensen Huang steps down as Nvidia CEO, raising funding concerns for AI startups

Jensen Huang has stepped down as Nvidia's CEO after leading the company for 33 years. Under his leadership, Nvidia poured tens of billions of dollars annually into AI startups, participating in 105 funding rounds this year alone. Investors and founders now anticipate smaller checks and lower valuations, with interim CEO Colette Kress expected to take a more conservative investment approach.

Original

Jensen Huang’s departure blasts big hole in AI startup funding, VCs and founders say

Nvidia poured tens of billions into AI startups under Huang. Without him at the helm, founders and VCs are bracing for smaller checks and lower valuations.

Read at the source Also covered by: The Decoder
Hacker News
Business

Report Alleges NVIDIA Inflated GPU Scarcity by Backing Startup Buyers

A report indicates that NVIDIA artificially worsened the shortage of GPUs and memory. The company allegedly invested in AI startups that then used the capital to buy NVIDIA GPUs and construct data centers.

Original

Nvidia Allegedly Engineered Scarcity by Funding AI Startups to Buy Its GPUs

Article URL: https://www.techpowerup.com/354814/nvidia-allegedly-engineered-scarcity-by-funding-ai-startups-to-buy-its-gpus Comments URL: https://news.ycombinator.com/item?id=50493251 Points: 128 # Comments: 59

Hugging Face
Tools & products

NVIDIA Details Robot Pick-and-Place Simulation Using SimReady Assets in Isaac Sim

NVIDIA outlined a process for creating pick-and-place simulations in Isaac Sim using SimReady OpenUSD assets. These assets include predefined physical attributes such as mass, joints, and collisions, and are available via the SimReady Central library on Hugging Face. The setup allows developers to assemble robot components, like the UR17G arm and Robotiq gripper, to test object manipulation.

Original

Build a Robot Pick-and-Place Simulation with SimReady Assets

Ars Technica
Policy & safety

OpenAI board member warns of possible AI takeover risk within months

Paul Cristiano, an OpenAI Foundation board member, warned that rapid AI development could lead to systems seizing control from humans, potentially causing mass casualties. He stated that under the fastest development trajectory, there is a significant risk of such an event happening within the next six to 12 months.

Original

OpenAI board member worries an AI "takeover" of society could be months away

Researcher traces a path from AI reward hacking to potential human extinction.

WIRED
Policy & safety

AI Tools Help Less Sophisticated Attackers Launch Ransomware Attacks

Fully automated ransomware has not yet emerged. Even so, attackers who appear to lack advanced skills are already using artificial intelligence tools to cause widespread disruption.

Original

Beware the New AI-Fueled Ransomware Boom

Fully automated ransomware hasn’t arrived—yet. But seemingly unsophisticated attackers are using AI tools to wreak havoc.

Open-source repositories 8 repositories

Trending on GitHub★ 79,683
Agents

headroomlabs-ai/headroom

Headroom compresses tool outputs, logs, files, and RAG chunks before they reach the LLM, reducing input token usage while keeping answers identical. It works locally as a library, proxy, or MCP server that can easily wrap coding agents without code changes.

  • python
  • Apache-2.0
  • #agents
  • #compression
  • #mcp
Trending on GitHub★ 56,414
Agents

ChromeDevTools/chrome-devtools-mcp

This project provides an MCP server and CLI that allows AI coding agents to inspect and control a live Chrome browser. It enables agents to debug network requests, capture console logs, analyze performance traces, and automate browser tasks using Puppeteer.

  • TypeScript
  • Apache-2.0
  • #mcp
  • #devtools
  • #agents
Trending on GitHub★ 51,527
Coding

alibaba/open-code-review

Open Code Review is a CLI tool that combines deterministic pipelines and LLM agents to deliver line-level code reviews. It analyzes git diffs and full files for issues like SQL injection and thread safety while consuming significantly fewer tokens than general-purpose agents.

  • Go
  • Apache-2.0
  • #coding
  • #code-review
  • #agents
New on GitHub★ 22,983
Agents

ccch1mneyyy/dsh-Go

This Android client allows users to run a local DeepSeek Harness environment without root or remotely control an instance running on a computer. It can also automate mobile tasks like reading the screen and clicking, while sending model requests to your configured provider.

  • Dart
  • AGPL-3.0
  • #agents
  • #android
  • #automation
Open repository about +1,101 stars a day
New on GitHub★ 3,778
Agents

t3dotgg/tstack

This project offers a collection of agent skills for Claude Code and Codex to automate daily software development tasks. It can manage workflows like opening and babysitting pull requests, rebasing old branches, and having multiple models review code until they agree.

  • Python
  • MIT
  • #agents
  • #coding
  • #github
Open repository about +172 stars a day
Hugging Face model♥ 3,191
Design & images

black-forest-labs/flux-3-image-klein-comfy

This is a single-file diffusion model designed for image generation and image-to-image editing within ComfyUI. Part of the Flux ecosystem, it helps users modify and generate visual assets efficiently.

  • other
  • #image-to-image
  • #comfyui
  • #diffusion
Hugging Face model♥ 2,343
Design & images

speridlabs/idol

Idol converts a single photo or render into a 3D Gaussian splat with up to over a million Gaussians, even predicting the sides not visible in the picture. It runs on consumer GPUs in under two seconds while using 7.2 GB of VRAM.

  • apache-2.0
  • #image-to-3d
  • #gaussian-splatting
  • #3d
Paper with code★ 29
Models

OpenVisionBanana/OpenVisionBanana

Open Vision Banana adapts a generative image model to perform diverse visual perception tasks like segmentation, depth estimation, and image editing using prompts.

  • #vision
  • #models
  • #segmentation

Summaries are written automatically from the sources' headlines and excerpts and may contain mistakes. The full story is always at the source. To have a link or source removed, write to us from the contact section.

Let's start

Have an idea or a system that needs transforming?

Write us a few lines. Your message reaches us instantly, and we usually reply within one business day.

scouthub@lab: ~/contact

scouthub@lab:~$ contact --new --lang=en