Because the resources required to run an overlord-level AI require a megacorp to fund them, and a megacorp to run the data centers (plural) needed to operate the neural models.
Open source models running on local hardware serve the needs of individuals and do what you want. But you're not going to be able to provide or gain access to much beyond what you can pay for. Megacorps like Microsoft, ABC, or Apple can pay for an awful lot.
Rods from God are actually... impractical. Controlling where the rods fall turns out to be much much harder than the idea makes it sound.
I suppose if you want non-nuclear ways to wipe out cities, sure. But it doesn't help if your target is much tighter, like military hardware or ships.
Personal reading: Models of the Mind by Grace Lindsay
Good take on the history and current state of modern neuroscience and its interconnectedness with modern machine learning and data science.
Reading to kiddo: Harry Potter and the Sorcerer's Stone by J.K. Rowling
I have a sore throat already from doing all the voices, what with all the shouting from Uncle Vernon and Hagrid in the first few chapters.
I write with the assumption that my journals will be read either when I'm not around, or after I die. Were I to write everything, I'd have to burn the books when I finish one, and there's stuff in them I'd prefer to keep.
Close enough to Harry's age in the first book. Old enough that we can successfully contextualize the bad behavior depicted in the book.