That is a noisy thread. His basic points:
- Revisit decision against using AOSP as the software; it would offer better privacy / security and wouldn’t require making an entirely new viable software stack
- be careful misinformation about AOSP is not propagated: it doesn’t require Google to do anything
- support Android and prioritize security features
- offer verified boot and A/B updates
- for security reasons, either ship firmware updates for nonfree components or do not use any
- microphone switch is a plus
- hardware support for virtualization is irrelevant
- CopperheadOS should not be run in containers
- harden the host OS
- desktop Linux stacks have weak areas: systemd, pulseaudio, tons of C and C++ at the application layer, no real application security model, permission model, no comparable full system SELinux policy…
- applications should not be trusted especially if installed through bleeding edge flatpak
- Android works fine on generic arm / arm64 hardware with mainline kernels; the frozen LTS branches are for out-of-tree drivers SoC drivers regardless of Android use
- provide an AOSP board support package and properly configured kernel
- get AOSP at least close to fully passing the Compatibility Test Suite on the L5