The end of feature discovery as we know it

(elenaverna.com)

2 points | by flail 3 hours ago ago

2 comments

  • bradknowles 3 hours ago ago

    If my only interface to a tool or program is a prompt box, then I’m leaving and never coming back.

    Having the UI present to me a coherent visual set of menus and options and tools is part of what helps inspire me to try new things. To think about what I’m trying to do in a different way. If you hide all that from me, then I have no idea of ways that I can experiment.

    I remember the old days of shell prompts. How you could do anything, if you wrote the script or program to do it. That’s okay for the part of any system that needs an ad hoc component. But it also takes a lot of training to use that CLI interface, and a lot of trial and error, and a lot of reading books and blogs. I think most people aren’t willing to put in that kind of work anymore.

    For me, I think the best way to integrate this kind of functionality is through whatever interface you provide for the user to get help. Have them describe the problem they’re having, or what they’re trying to do. That is the right time to expose any additional functionality you have that can help them do what they want. Have a simple and easy-to-use starting point, and then only expose the additional functionality when it is needed or desired.

    • flail 2 hours ago ago

      I think, fundamentally, we're going to have a divergence of UI. Some products will bet heavily on AI agents, and they won't care about UX in the "classic" sense. Others would still target humans. For those, not that much has changed.

      One product we work on is a perfect example. The founder went for ChatGPT-like interface. The early adopters want the old-school filters, so they can find the interesting stuff in a way they're used to.

      A follow-up consideration is that, right now, an AI agent can be a user. It isn't necessarily capable of being a customer: https://pawelbrodzinski.substack.com/p/an-ai-agent-is-not-yo...

      Overreliance on agentic usage may be tricky from that perspective.