Lumo isn’t really private. If you’re not local, you’re not private. AFAIK it sends to the model in cleartext (zero access) which isn’t E2EE. Signal, for example, is E2EE. Based on internet reviews, you’re better off running the models they offer locally for performance reasons on the cost anyway. I’d say you’re better off without AI in almost every situation you’d want to use an LLM of such such limited capability too so unless you really really want to set resources on fire it’s better to just drop it period.

Netflix cannot be replaced by Jellyfin. That is very disingenuous. To “replace” Netflix with Jellyfin requires a bit more infrastructure and while it might be cheaper there’s risk involved that isn’t present with your other changes.