Creators are finally fighting back on AI training

Japanese content creators and their representatives are demanding OpenAI stop using their copyrighted works to train AI models, reflecting growing global friction over how generative AI companies obtain training data and the ease with which users can now recreate copyrighted styles and characters.

TechCrunch

Shopify is experiencing explosive growth in AI-driven shopping, with AI traffic up 7x and orders up 11x, signaling a major shift toward agentic commerce as the company positions AI as central to its platform strategy.

Thread Reader

Vector search, despite widespread adoption, has fundamental flaws in production retrieval compared to classical BM25, while Kimi K2's transparent training report offers rare insights into frontier model development methodology and scaling approaches.

Sebastian Raschka's Magazine

Beyond standard LLMs, emerging architectures like linear attention hybrids, text diffusion models, code world models, and small recursive transformers each offer distinct trade-offs—trading efficiency for capability or specialization for generalization—while traditional transformers remain proven and optimal for most current applications.

Alibaba

Alibaba's Qwen3-Max-Thinking early preview achieves perfect scores on elite reasoning benchmarks while still in training, signaling major progress in Chinese LLM competition and test-time compute scaling.