These are personal notes on how gen AI impacts software applications. This is a follow-up to this post. What gen AI can do as of August 2025 1. Take multimodal files as input Until last year, most gen AI apps were limited in the kinds of files they could read. Gemini couldn’t even read PDFs if I recall correctly. This barrier has fallen. Most apps now accept txt, csv, pdf, docx, xlsx, pptx, any xml or json files. More importantly, all major gen AI apps are now multimodal: they accept text b...
