Native SDK - Toolkit for building native desktop apps

The "no browser, no WebView, no compromise" positioning is very clear.

Desktop apps are in a weird place where a lot of tools feel like websites wrapped in a window, which is fine until performance, memory, offline behavior, or native feel actually matter. A toolkit that gives declarative UI and a predictable state model while still rendering natively sounds like a strong direction.

Curious how close Native SDK gets to platform-native behavior out of the box. Does it handle things like keyboard shortcuts, menus, window controls, accessibility, and OS-specific patterns automatically, or do developers need to wire a lot of that themselves?

How does pricing scale once you start getting serious traffic — is there a clear breaking point where it jumps from a flat tier to something unexpectedly higher?

A built-in cost dashboard showing real-time bandwidth and function invocation spend per project would be huge. Right now I have to dig through usage logs and try to estimate monthly bills myself, which makes budgeting for client work way harder than it needs to be.

I've been driving native desktop apps with automation and half my breakage comes from each OS handling windows and menus differently. Does Native SDK give you one API across Mac and Windows, or are you still writing platform-specific code under the hood for the fiddly stuff like tray icons and file dialogs? That gap is usually where cross-platform toolkits quietly leak.

Usually going native like that means giving up easy stuff like embedding web content or styling with CSS. Is the tradeoff worth it in practice, and what's it actually like to build with something closer to React, or something new to learn? Excited to try it out!

How does pricing actually work for mid-sized projects, and are there any hidden limits on bandwidth or build minutes I should know about before committing?

I am excited about this - having an alternative to the standard Electron/Tauri and having explored Flutter and React Native, this sounds very compelling and the sample apps - especially the note-taker - look quite smooth. I am wondering if the fact that our platform is a monorepo with React/Next/Tailwind/TypeScript will be a multiplier in being able to use this? We've built heavily on Vercel SDKs (we use the AI Gateway, the AI SDK, Nextjs) and we're big V0 users as well! Love this, @Vercel!

I thought this kind of approach was already outdated. Does it still make sense to have this type of hybrid platform in the world of AI assistants, where building natively is no longer a big challenge? And you no longer need to rely on third parties to use recent native APIs or to avoid surprises when your packages and app break.

How does pricing work once you go beyond the hobby tier — is it purely based on bandwidth and build minutes, or do serverless invocations factor in too? Trying to estimate what a mid-sized Next.js app would realistically run.