the-decoder

NYU professor fights AI cheating with AI-powered oral exams that cost 42 cents per student

Panos Ipeirotis, a professor at NYU Stern School of Business, used an AI voice agent built on ElevenLabs to orally examine 36 students after discovering many couldn't explain their own polished ...
the-decoder

ByteDance's StoryMem gives AI video models a memory so characters stop shapeshifting between scenes

Researchers from ByteDance and Nanyang Technological University have developed StoryMem, a system designed to maintain visual consistency in AI-generated videos across multiple scenes. The system ...
the-decoder

Project Gumdrop: OpenAI's first AI gadget could send handwritten notes to ChatGPT

OpenAI has reportedly shifted production of its first AI hardware device from Luxshare to Foxconn. The company wants to avoid manufacturing in China, according to Taiwanese business newspaper Economic ...
the-decoder

LLMs could serve as world models for training AI agents, study finds

An international team of researchers has tested whether large language models can serve as world models: internal simulators that predict outcomes when an AI agent takes a specific action. After ...
the-decoder

OpenAI seeks new "Head of Preparedness" for AI risks like cyberattacks and mental health

OpenAI is hiring a Head of Preparedness. The position focuses on safety risks posed by AI models. OpenAI CEO Sam Altman points to the now well-documented effects of AI models on mental health as one ...
the-decoder

OpenAI insists its shopping suggestions shouldn't be seen as advertising

Paid users of ChatGPT recently saw a prompt asking them to connect their account with Target for shopping, which some interpreted as an advertisement. But an OpenAI product manager denied that any ...
the-decoder

Chatbots are now rivaling social networks as a core layer of internet infrastructure

Generative AI systems reached around seven billion monthly web visits in September 2025, a 76% increase compared to the previous year, according to Similarweb. ChatGPT remains at the center of the AI ...
the-decoder

Qwen3-VL can scan two-hour videos and pinpoint nearly every detail

Alibaba's Qwen3-VL, launched in September, outperforms GPT-5 and Gemini 2.5 Pro on benchmarks that require solving math questions using images, analyzing videos, and understanding documents. The ...
the-decoder

DeepseekMath-V2 is Deepseek's latest attempt to pop the US AI bubble

Chinese AI start-up Deepseek has introduced DeepseekMath-V2, an AI model that achieves gold medal-level results in the International Mathematical Olympiad (IMO). The model can evaluate and iteratively ...
the-decoder

As Google pulls ahead, OpenAI's comeback plan is codenamed 'Shallotpeat'

According to an internal memo from The Information, OpenAI CEO Sam Altman warned staff that Google's recent AI progress with Gemini 3 could create "temporary economic headwinds" for the company, as ...
the-decoder

OpenAI could face a billion-dollar fine over claims it used pirated books in AI training

OpenAI could soon face a billion-dollar fine. Authors and publishers suing the company for copyright infringement have uncovered internal messages and emails about the deletion of a dataset containing ...
the-decoder

Google boasts 1.3 quadrillion tokens each month, but the figure is mostly window dressing

Google reports processing over 1.3 quadrillion tokens per month with its AI products, a sharp increase since May that is likely driven by the adoption of more complex reasoning models like Gemini 2.5 ...