Can you explain how this is better than Google End-to-End? (Which isn't completed, I know.)
Or even how it's worth implementing over regular webmail in the first place?
Extensions are obviously not optimal, for the very same reason crypto in JavaScript is not good: timing attacks, cache attacks, optimizations, secure storage (in-memory and on-disk), random-number generation, verifiability, ....
Protonmail suffers from that too. Except it's also subject to key exfiltration. Without an exploit. The server simply promises not to send malicious JavaScript.
It isn't significantly more involved for an attacker who has breached their servers to send malevolent code. It's also not easy to detect.
> We will never have a useable experience for all user if we do not accept compromises.
There are usable end-user interfaces capable of featuring secure cryptography.
Desktop applications. Mobile apps. Browser extensions. Hardware tokens (U2F).
Those can be audited. Inspected. Users can freeze updates.
Trying to implement secure cryptography in JavaScript that is fetched repeatedly from an assailable server is just asking for trouble.
> GPG is already not perfect and we should move away from it anyway
Yes, but that doesn't mean ignoring the advancements that have been made in modern security.
The most appropriate response to securing email is still this: Don't, use a secure messenger; or use out-of-band encryption. It's likely to remain that way forever.