Nvidia will use NitroGen to create universal AI agents for all worlds.Anthropic's AI model solves tasks with a time horizon of almost five hours.Alibaba's new AI model makes static images editable after the fact.Semi-open AI models from China are displacing the US competition.
The AI news for December 22nd, 2025
--- This episode is sponsored by ---
Find our more about our today's sponsor Pickert at pickert.de.
---
Would you like to create your own AI-generated and 100% automated podcast on your chosen topic? --> Reach out to us, and we’ll make it happen.
Here are the details of the day's selected top stories:
Nvidia will use NitroGen to create universal AI agents for all worlds.
Source: https://the-decoder.de/nvidia-will-mit-nitrogen-universelle-ki-agenten-fuer-alle-welten-erschaffen/
Why did we choose this article?
NitroGen is a large-scale vision-action foundation model trained on 40,000 hours of gameplay — a concrete step toward general-purpose interactive agents. Useful for listeners tracking how game-trained models can transfer to real-time simulation, robotics, and AI-driven interactive experiences.
Anthropic's AI model solves tasks with a time horizon of almost five hours.
Source: https://the-decoder.de/ki-modell-von-anthropic-loest-aufgaben-mit-fast-fuenf-stunden-zeithorizont/
Why did we choose this article?
METR's test showing Claude Opus 4.5 handling ~4h49m horizons signals progress in long-context planning and sustained reasoning. Important for anyone building agents that must manage extended tasks, workflows, or multi-step coordination without frequent human resets.
Alibaba's new AI model makes static images editable after the fact.
Source: https://the-decoder.de/neues-ki-modell-von-alibaba-macht-statische-bilder-nachtraeglich-editierbar/
Why did we choose this article?
Qwen-Image-Layered's RGBA layer decomposition makes non-destructive, precise edits of static photos possible — a practical advance for creators, game devs, and localization workflows. Signals how generative tools are moving toward fine-grained content control rather than blunt replacement.
Semi-open AI models from China are displacing the US competition.
Source: https://the-decoder.de/halb-offene-ki-modelle-aus-china-verdraengen-die-us-konkurrenz/
Why did we choose this article?
A shift in downloads toward Chinese 'half-open' models is a strategic indicator of ecosystem and adoption changes. This matters for researchers, policymakers, and businesses assessing where innovation, talent, and practical deployments are concentrating.
Do you have any questions, comments, or suggestions for improvement? We welcome your feedback at podcast@pickert.de.
Published on 3 days, 20 hours ago
If you like Podbriefly.com, please consider donating to support the ongoing development.
Donate