Now you might ask: what about AI native applications? Well, as it turns out, most of them are essentially clones of existing software but with a chatbot slapped on top. Because of the error prone nature of AI, any application that leverages it also has to surface all of the controls required to override all of its decisions. So you end up with a traditional software application plus AI.
AI promised to transform and even replace software applications, but all it did instead was augment them with an unreliable chatbot. All of the old fields and buttons are still there, but now there’s an additional text field that you can type into and hope for the best.
GT often uses the wrong word or changes the tone of a message. AI always gets the intent right and always seems to use the most appropriate words given the original intent/meaning (or at least something way better than what GT does). And, whenever there is doubt, I can argue with it, so AI is happy to explain the nuances and differences between the possibilities.
Edit: I recently had to send a semi-formal email requesting something from a government employee in a different country (using a language I'm a beginner at), and AI was immensely helpful in getting the right tone (neither informal or too formal) and everything else right. The Google Translate version of what I had originally written was miles and miles and miles worse than what AI helped me craft.
Refactoring tools in my IDE: In some cases where I could use the refactoring tools in my IDE I will ask the assistant to do something for me instead, of course it will also make changes that the refactoring tools won't do such as tear apart a complicated if-then-else ladder.
Photo retouching: there are plenty of photo retouching jobs that can be done easily with AI or with the tools built into Photoshop, which one is better depends on the situation. I really wish I had an AI tool to make masks ("cut out the person") that I could then use with the other tools or with AI generation.
"slapping on a chatbot" is a v0 attempt at re-imagining what software looks like. It's not very inventive and sometimes it sucks, but it's easy to understand and implement and we're very early in this era.
The distribution of change also isn't uniform. Excel might not have changed dramatically, but software engineering apps are evolving rapidly. Clawdbot/moltbot hint at new forms of personal computing. Look for the future where the optimism is.