Better to build full native but rely on cross-platform architecture, design patterns, common backend/infrastructure, etc than settle for hypebeast cross-platform abstraction magic with dependencies on overburdened and leaky by definition integrations back to native SDK. The former yields all the proven benefits of native apps and strong platform vendor support with a contingency plan if said ecosystem launches a competing product. The latter leads to a world of hurt.
Specifically, the article misses the mark wrt Swift. Swift is open source and truly cross-platform. SwiftNIO, Tensorflow, and the server-side Swift projects are a great example of how far one can go assuming a willingness to invest in the language completely devoid of any Apple API involvement.
It truly seems like the next major breakthrough would be if SwiftUI could (relatively easily) be implemented for desktop Windows or *nix. Sure, the article notes people tried the same thing with GNUStep or WinObjC, but you were always truly tied to or ended up missing Cocoa, Foundation, and the Obj-C runtime. Conversely & ironically, swift-evolution, spm, and many of the most "Swifty" parts of the Swift ecosystem come to the other toolchains long before they make it into an annual Xcode release.