How to build a billion dollar audio app in a weekend

Zoom also uses dark patterns to try to get you to install the desktop app, even if you could join via browser.

Zoom's browser client doesn't use the browser's underlying WebRTC implementation for encoding/decoding and bandwidth management. Instead, Zoom cross-compliled their proprietary h264 and network stack to web assembly, and that does the heavy lifting.

Which is a really, really cool hack. But the wasm instruction set is much more limited than what you can do in natively compiled code, and access to the browser's networking internals are similarly constrained.

The Zoom web client doesn't -- and probably won't ever, with this architecture -- be able to do a good job with calls with more than 4 or 5 participants, or be able to handle the full range of real-world network conditions that are necessary to achieve what I'd call "just fine" for broad use cases.

Conversely, Zoom's desktop client defines call quality for all of the rest of us working on this stuff. Our goal at Daily is explicitly to get to as good as Zoom, for all use cases, running in the web browser. For "normal" video calls up to about 12 people, we're there now. For larger calls, especially on low-end hardware, there's still work to do.