PWAs could have been so good. redbean/llamafile might be the closest, though.

You can lose the ugly "exit" at the end and space it right for formatting too

    // 2>/dev/null; exec go run "$0" "$@"

Then leave and wait for it to start sending us poetry about its constant mortal fear of the night :)

[1] https://youtube.com/watch?v=4PUIxEWmsvI

If you’re permitting desktop apps, https://github.com/diamondburned/gtkcord4 is an option.

The built in USB-C ports are actually fully compatible with Thunderbolt they're just waiting for certification before they can say that. https://community.frame.work/t/thunderbolt-yes-or-no-or-not-...

This is correct; The customization flow advertises the modules as USB4, which is a superset of Thunderbolt 3's capabilities (which itself was a superset of USB 3).

You can usually expect them to work as supersets. But I believe it's possible to implement USB4 without all the Thunderbolt compatibility, and Thunderbolt 3 has some situations where it's not backwards compatible with USB 3.

AFAIK USB4 now mandates Thunderbolt's Type C altmode.

That was Intel's concession on blocking PCI-E 4.0: they wanted to mandate Thunderbolt over Type C as the official external PCI-E interface in enterprise, when the entire industry went with OCuLink instead, due to actually being enterprise-grade (it uses the replacement of the SFF-8644 (multiuse, commonly called MiniSAS HD), the SFF-8611; some mobos use SFF-8644, and SF-8644<->SFF-8611 cables exist) whereas Type C is not.

When they got their way with USB4 mandating Thunderbolt, PCI-E 4.0 finally was ratified several years late.... PCI-E 4.0 is now so late because of Intel's ridiculous tantrum that only two generations of PC will use it, and the world is proceeding directly to 5.0. Some products have already only exist in 3.0 and (soon to be) 5.0, skipping 4.0 entirely.