Subscribe

Shane

Dec 24, 2025Dec 24, 2025

Category

Empty

1.

MIT Technology Review reported that social media posts had misrepresented claims that GPT-5 solved multiple unsolved mathematical problems, and noted that the startup Axiom later reported AxiomProver had solved two Erdős problems and performed strongly on the Putnam competition.

2.

OpenAI admitted that prompt injection attacks might never be fully solved and stated that it was using automated red teaming to mitigate prompt-injection risks in ChatGPT Atlas.

3.

Alibaba Cloud released two new Qwen voice models that could create or clone voices from three seconds of audio using text commands.

4.

Zhipu AI introduced GLM-4.7, a low-cost model specialized in autonomous programming that used a "Preserved Thinking" approach to retain reasoning across long conversations.

References

1.

https://www.technologyreview.com/2025/12/23/1130393/how-social-media-encourages-the-worst-of-ai-boosterism/

How social media encourages the worst of AI boosterism

The era of hype first, think later.

technologyreview.com

1.

https://the-decoder.com/openai-admits-prompt-injection-may-never-be-fully-solved-casting-doubt-on-the-agentic-ai-vision/

OpenAI admits prompt injection may never be fully solved, casting doubt on the agentic AI vision

OpenAI is using automated red teaming to fight prompt injections in ChatGPT Atlas. The company compares the problem to online fraud against humans, a framing that downplays a technical flaw that could slow the rise of the agentic web.

the-decoder.com

1.

https://the-decoder.com/alibabas-new-qwen-models-can-clone-voices-from-three-seconds-of-audio/

Alibaba's new Qwen models can clone voices from three seconds of audio

The Qwen team at Alibaba Cloud has released two new AI models that create or clone voices using text commands.

the-decoder.com

1.

https://the-decoder.com/zhipu-ai-challenges-western-rivals-with-low-cost-glm-4-7/

Zhipu AI challenges Western rivals with low-cost GLM-4.7

Zhipu AI has introduced GLM-4.7, a new model specialized in autonomous programming that uses "Preserved Thinking" to retain reasoning across long conversations.

the-decoder.com

more pain more gain 🚀

© 2024-2025 Shane "Lx". All rights reserved.