It depends on the type of facts, but sometimes it’s much easier to verify an answer than to get the answer in the first place. For example sometimes the LLM will mention a keyword that you didn’t know or didn’t remember and that makes googling much easier.

And even among runtime environments some are much better than others. I don’t really mind dotNet or the JVM that much, but Electron seems particularly wasteful. I don’t use a single Electron app.

How could I forget to mention ctrl+f!

There are a couple of functions that web apps almost always have and that native apps tend to lack: (1) selecting and copying text from anywhere in the app to the clipboard; (2) bookmarking individual views within the app. Of course, natives apps in principle could be faster and use more of your hardware —in practice though, they tend to be horribly bloated electron crapps. 😅 So yeah, a decent native app can be better than a web app, but good luck finding one for your purpose.

One interesting feature in this paper is that the programmers who used LLMs thought they were faster, they estimated it was saving about 20% of the time it would have taken without LLMs. I think that’s a clear sign that you shouldn’t trust your gut about how much time LLMs save you, you should definitely try to measure it.

Normally people use ChatGPT to vibe code, this is the first instance I’m aware of of ChatGPT using people to vibe code!