Both protocols solve the same problem: making proxy traffic indistinguishable from a normal visit to an ordinary website. The censor sees a TLS connection to some host, nothing more. The difference is in how they get there.
Trojan: a good lie needs a real website
Trojan connects to a real, existing HTTPS site — the proxy borrows its domain and certificate. To anyone watching, your traffic is a session with that site. If the proxy is probed directly, the connection behaves like the ordinary website would.
The cost: you need a domain and a certificate, and the disguise lives or dies with the site it imitates.
VLESS: a lighter inner protocol
VLESS is a minimal protocol for the channel itself. It carries almost no overhead, supports modern transports — including Reality, which borrows the fingerprint of someone else’s TLS handshake without owning any certificate at all. To the censor the connection looks like traffic to a large popular service; there is no domain of yours to block.
VLESS also has first-class support in mobile and desktop clients, which is why one subscription link works everywhere.
Why we run both
Any protocol can fall out of favor with a censor in a specific country or on a specific network. Two protocols on every node means a fallback that does not require a second purchase: switch the config in your app and keep working. The subscription link carries both — the app picks what connects best.