💯 it's pretentious and unproductive. Your setup should be threat model specific. Don't let perfect be the enemy of good.
Discussion
To a point, but let's not kid ourselves here. When the enemy is in complete collusion (big tech, state intelligence and law enforcement) we can't afford to draw arbitrary lines between which institutions we're defending against. We have to go as far as possible. Don't let perfect be the enemy of good, but if imperfection means leaving leaks in your chain of data protection, it's not doing much good.
For instance, using Simplex or Signal on iOS doesnt do much good when Apple holds the keys to encryption remotely and you can't verify their code. In this case there's not much benefit compared to iMessage since you're still at the mercy of Apple (and any secret court subpoenas or state intelligence backdoors which certainly exist)
We must study what serious threat actors like hackers or darknet vendors use for communication and online activities and adopt accordingly. These actors require millions of state intelligence funding just to take them down. Imagine if the average American, conducting no crime, requires this level of investigation just to get simple metadata. The surveillance state would be dead overnight. This is the goal.
Using the Tor Browser on Linux on a coreboot computer is very easy these days and costs as much as any other computer.
Using only FOSS apps on a GrapheneOS pixel has a initial learning curve, but after is just as easy as stock pixel android or a Samsung.
Combine this with nostr:npub18dlusgmprudw46nracaldxe9hz4pdmrws8g6lsusy6qglcv5x48s0lh8x3's advice of maintaining general cognisance of privacy in public and you end up with a public that is immune from even state level intelligence.
Why should we settle for less when modern Linux and open source android offers so much in usability and productivity compared to windows and iOS?