AI video models

Wired reports on a new wave of AI research that trains models on how real people play video games, including their mistakes and sloppy habits, rather than only on expert or scripted play. The idea is that human error contains useful data for teaching AI systems to behave more naturally. The piece is being shared and discussed in tech and gaming circles.

Kernel developer Greg Kroah-Hartman, the maintainer of the Linux kernel stable branches, has given a talk on what large language models mean for software security. The presentation examines how AI-generated code affects vulnerability handling and maintenance work in large open source projects. The talk is circulating among developers and technology commentators, with early responses still limited but interest growing in how core infrastructure maintainers view LLM-driven risks.

Kernel maintainer Greg Kroah-Hartman discusses software security in an era when large language models are increasingly used to write code. The talk, shared as a video, examines how AI-generated contributions affect the Linux kernel's review process and the challenges of maintaining security standards as LLM-assisted development spreads. Readers are weighing how maintainers can vet code at scale when machine-written patches grow in volume.

A British startup is turning video game controller inputs into training data for AI models designed to navigate the physical world. The idea is that human gameplay, including mistakes and clumsy manoeuvring, offers rich examples of how people learn to control complex systems, which could help robots and other embodied AI improve. The approach is drawing attention as a cheaper alternative to real-world data collection.

A new project called agent-wow showcases GPT-6 Astra, described as an AI agent, playing World of Warcraft for the first time. The demo is drawing attention as an example of AI agents handling complex, open-ended game environments, with people debating how far large language model agents have come at operating software meant for humans.

Meituan, the Chinese food delivery and services giant, has released LongCat-Video, a new open-source AI video generation model on GitHub. The project, written in Python, is drawing developer attention and climbing repository trending charts globally, as the company continues expanding its LongCat AI lineup beyond language models into generative video.

MoneyPrinterTurbo, an open-source Python project by developer harry0703, automates the creation of high-definition short videos. Users enter a topic or keyword and the tool uses large AI models and an automated workflow to produce a finished clip. The project is gaining attention among developers and content creators interested in AI-powered video generation.

A technology reviewer has tested LTX-2.5, examining its potential for physical AI applications including robotics, world models and synthetic data generation. The assessment highlights how generative video models could help train robots by simulating realistic environments. Interest in LTX-2.5 is growing among developers exploring simulation-driven approaches to embodied AI, where real-world training data is expensive and difficult to collect at scale.

AI video company Tavus has introduced Griffin, a conversational video model it says is realistic enough to pass a video Turing test. The announcement is drawing attention to how closely AI-generated humans now resemble real people on camera, and to what that means for trust in video, digital identity and online deception.

Runway has introduced Praxis-1, a new world action model designed for robotics. Announced via The Robot Report, the model marks the AI video company's move into robotics, suggesting a system aimed at helping robots understand and act in physical environments. Details on capabilities and partners were not immediately available.

US startup Tavus has unveiled Griffin, a conversational video model designed to power lifelike AI avatars on video calls. In company testing, 48% of participants mistook the AI-generated call partner for a real human, a result being cited as evidence of how quickly synthetic video is approaching human indistinguishability and raising fresh questions about deception and consent in video communication.

A developer is demonstrating Xiaomi's MiMo V2.6 Qwen 9B distill model running locally on a 16GB MacBook Pro, the third installment in a tutorial series launched this month. The series covers AI, programming and agents, with an emphasis on inclusive, community-oriented learning. Interest centres on whether mid-range laptops without high-end GPUs can now run capable open-weight language models.