Comment by vintagedave
4 hours ago
> If "just build it natively" were actually easier, faster, or cheaper at scale, everyone would do just that
Value prop of product quality aside, isn't the AI claim that it helps you be more productive? I would expect that OpenAI would run multiple frontends and that they'd use Codex to do it.
Ie are they using their own AI (I would assume it's semi-vibe-coded) to just get out a new product or using AI to create a new product using the productivity gains to let them produce higher quality?
It's about consistency - you want to build an app that looks and functions the same on all platforms as much as possible. Regardless of if you are hand-coding or vibe-coding 3 entirely separate software stacks, getting everything consistent is going to be a challenge and subtle inconsistencies will sneak in.
It comes back to fundamental programming guidelines like DRY (Don't Repeat Yourself) - if you have three separate implementations in different languages for everything, changes will be come harder and you will move slower. These golden guidelines still stand in a vibe-code world.
On a side note, the company I work for (RemObjects, not speaking on their behalf) has a value ethos specifically about using the native UI layers, and encouraging our customers to do the same. (We make dev tools, a compiler supporting six languages (C#, Java, Go, etc) plus IDEs.)
Our IDE does this: common code / logic, then a native macOS layer and a WPF layer. Yes, it takes a little more work (less than you'd think!) but we think it is the right way to do it.
And what I hope is that AI will let people do the same -- lower the cost and effort to do things like this. If Electron was used because it was a cheap way to get cross-platform apps out, AI should now be the same layer, the same intermediate 'get stuff done' layer, but done better. And I don't think this prevents doing things faster because AI can work in parallel. Instead of one agent to update the frontend, you have two to update both frontends, you know?
We're building an AI agent, btw. Initially targeting Delphi, which is a third party's product we try to support and provide modern solutions for. We'll be adding support for our own toolchains too.
What I fear is that people will apply AI at the wrong level. That they'll produce the same things, but faster: not the same things, but better (and faster.)