I've spent my career designing and building systems to help humans understand data and control computers. While I find this hard to swallow it's also hard to argue with. Tens of thousands of engineers and designers rebuilding slightly different drop-downs is an inefficient world that is unarguably coming to an end.
As so much of the first-line decision-making moves to LLMs there's definitely going to be opportunities for much richer and complex output from LLMs - how we can create terse and expressive visual summarisations/interfaces for where humans need to make decisions. But it's a much smaller world.
Where I suspect the wheels are going to come off for some though is that it's far, far easier to create a complex, difficult to understand UI than a simple one. And if simplicity and clarity are what enables effective LLM utilisation attempting to skip all that bothersome UX work will go poorly.
>Tens of thousands of engineers and designers rebuilding slightly different drop-downs is an inefficient world that is unarguably coming to an end.
As long as branding is important, you're going to keep getting slightly different dropdowns. In fact, you could argue that in a front end world that's dominated by AI written code that pulls from standard libraries, branding and all its different dropdowns become more important than ever.
As long as users are in the loop, interfaces will be important. The future of computer interactions would be greatly constrained if chat boxes were exclusive paradigm!
"Look at Cursor — it started as a full-blown IDE and is now converging on what's essentially a task list."
IDE functionality wasn't removed at all, and is a distinguishing factor from CLI and other tooling!
A real question - who or what are we building software for? If we're using AI to build apps that are just used by other AIs, then yes, why does anything need a UI?
But while AIs are doing all the creating and all the consuming, what exactly are us humans doing all day? Do we really think software for humans will shrink to just an Agent interface?
Jakob Nielsen has been banging the drum that we're going all end up with custom interfaces that will be tailored to what our current task is at that time. How this will work is TBD, but if I recall correctly, he's basically agreeing with your take around agents as an interface.
The idea that people are going to YOLO changes to DNS and Postgres migrations gives me such anxiety, knowing the pain people are in for when they “point Claude at it, one prompt, and done”, then their business is dead in the water for a week or two while every executive is trying to micromanage the recovery.
I love Streamlit and mermaid, but if these are the shining examples this isn’t a good sign. These have hard ceilings and there’s only so much you can work around the model of “rerun the entire Python script every time the page changes”.
As long as humans are involved the UI will matter. Maybe the importance is not on the end-user facing UI, and maybe it’s more on the backend SRE-level observability UI that gives insight into whether the wheels are still on the bus, but it will matter.
Some people are getting the AI to handle that too, and like all demos, that will work until it doesn’t. I’m sure NASA or someone can engineer it well enough to work, but it’s always going to be a tradeoff: how fast you can go depends more on the magnitude of the crash you can survive, than the top speed someone achieves once and lives to tell about it.
I've spent my career designing and building systems to help humans understand data and control computers. While I find this hard to swallow it's also hard to argue with. Tens of thousands of engineers and designers rebuilding slightly different drop-downs is an inefficient world that is unarguably coming to an end.
As so much of the first-line decision-making moves to LLMs there's definitely going to be opportunities for much richer and complex output from LLMs - how we can create terse and expressive visual summarisations/interfaces for where humans need to make decisions. But it's a much smaller world.
Where I suspect the wheels are going to come off for some though is that it's far, far easier to create a complex, difficult to understand UI than a simple one. And if simplicity and clarity are what enables effective LLM utilisation attempting to skip all that bothersome UX work will go poorly.
>Tens of thousands of engineers and designers rebuilding slightly different drop-downs is an inefficient world that is unarguably coming to an end.
As long as branding is important, you're going to keep getting slightly different dropdowns. In fact, you could argue that in a front end world that's dominated by AI written code that pulls from standard libraries, branding and all its different dropdowns become more important than ever.
As long as users are in the loop, interfaces will be important. The future of computer interactions would be greatly constrained if chat boxes were exclusive paradigm!
"Look at Cursor — it started as a full-blown IDE and is now converging on what's essentially a task list."
IDE functionality wasn't removed at all, and is a distinguishing factor from CLI and other tooling!
A real question - who or what are we building software for? If we're using AI to build apps that are just used by other AIs, then yes, why does anything need a UI?
But while AIs are doing all the creating and all the consuming, what exactly are us humans doing all day? Do we really think software for humans will shrink to just an Agent interface?
Jakob Nielsen has been banging the drum that we're going all end up with custom interfaces that will be tailored to what our current task is at that time. How this will work is TBD, but if I recall correctly, he's basically agreeing with your take around agents as an interface.
I still don't understand this idea. A different interface every time would be the highest cognitive load imaginable.
The idea that people are going to YOLO changes to DNS and Postgres migrations gives me such anxiety, knowing the pain people are in for when they “point Claude at it, one prompt, and done”, then their business is dead in the water for a week or two while every executive is trying to micromanage the recovery.
I love Streamlit and mermaid, but if these are the shining examples this isn’t a good sign. These have hard ceilings and there’s only so much you can work around the model of “rerun the entire Python script every time the page changes”.
As long as humans are involved the UI will matter. Maybe the importance is not on the end-user facing UI, and maybe it’s more on the backend SRE-level observability UI that gives insight into whether the wheels are still on the bus, but it will matter.
Some people are getting the AI to handle that too, and like all demos, that will work until it doesn’t. I’m sure NASA or someone can engineer it well enough to work, but it’s always going to be a tradeoff: how fast you can go depends more on the magnitude of the crash you can survive, than the top speed someone achieves once and lives to tell about it.