This week in AI — 2 stories

This week's stories share a quiet but important throughline: the AI field is increasingly targeting root causes rather than surface symptoms. CreativityNeuro goes straight to the weights to fix why language models think in circles, while OpenAI and Broadcom's Jalapeno chip tackles the infrastructure dependency that's constrained how and where large models actually run. Together, they're a useful reminder that the most consequential AI work right now isn't happening at the prompt layer - it's happening deep in the stack.