GLM-4.7 Retains Thought Across Dialogues. Three-Second Voice Cloning. OpenAI Blocks Prompt Injections. AI Breaks Smart Home Promise.

Show notes

The AI news for December 24th, 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:

The AI coding model GLM-4.7 remembers trains of thought across long dialogues.
Source: https://the-decoder.de/ki-coding-modell-glm-4-7-merkt-sich-gedankengaenge-ueber-lange-dialoge/
Why did we choose this article?
Introduces 'Preserved Thinking' for keeping chain-of-thought across long coding dialogs — a tangible technical advance for coding assistants with clear practical implications for reliability and autonomous programming.

Voice cloning from three seconds of audio: Qwen releases new TTS models.
Source: https://the-decoder.de/stimmen-klonen-aus-drei-sekunden-audio-qwen-veroeffentlicht-neue-tts-modelle/
Why did we choose this article?
Alibaba's Qwen TTS claims voice cloning from three seconds of audio — a significant leap for voice synthesis with direct consequences for content creation, authentication, and misuse mitigation strategies.

OpenAI resists prompt injections that threaten the 'agentic Web'.
Source: https://the-decoder.de/openai-wehrt-sich-gegen-prompt-injections-die-das-agentische-web-bedrohen/
Why did we choose this article?
Covers OpenAI's automated Red Teaming and the broader threat of prompt injections to agentic systems — essential reading for developers and product teams building agents and web-integrated AI.

How AI broke the smart home in 2025
Source: https://www.theverge.com/tech/845958/ai-smart-home-broken
Why did we choose this article?
A concrete, user-facing example of generative AI causing real reliability and UX failures in consumer devices — useful cautionary lessons for product teams, integrators, and users planning AI upgrades.

Do you have any questions, comments, or suggestions for improvement? We welcome your feedback at podcast@pickert.de.