Creative/weird AI experiments & early research [developing]

Physical AI hackathon panda robot (ReachyMini) from Hugging Face; Board Mode for agent sessions (Emad Mostaque) turning linear workflows into creative workspace; Stanford HAI new scaling law method cuts compute for one-shot LLM training, democratizing experimentation; AutoResearch AI survey maps research automation spectrum from human-steered to autonomous AI-led systems. Signals growing interest in agentic scientific discovery and playful robotics.