Deep research got significantly better since GPT 5.2 Especially while doing research and asking for multi step checklist with a handful variables
Always use a prompt engineering tool I don’t understand how people use LLMs without one
70