I am not a futurist

10+ years ago, I thought the future of app complexity could be solved by having text- and therefore also speech-based interfaces. I built a concept of an app that used this as a way of interacting with multiple social networks and news sites.

5 years ago, I thought the vast array of partially overlapping functionality in the multitude of apps could be addressed by having apps that offered extensibility points to allow different apps to provide specific functionality that was all consumed from one (or a few) place(s).

I couldn't persuade people that a text-based interface as the primary way of interacting with complex apps was the future. Nor were they convinced that my prototype could scale to more complex and varied input.

Rather than offer extensibility, app developers wanted to "own it all" and do everything themselves. The dream of controlling (& therefore monetising) everything was too appealing.

Now we're told that "Agentic AI is the future." Apps will all become agents, and we'll do everything through a chat interface.

Obviously, I see the parallels with my examples above.
But I don't see that as the one scenario to rule them all.

The use of "Agents" in AI/LLMs (probably connected via MCP) definitely has a place in the future and addresses some of the scenarios that AI (and even possible future General AI) could never do.

However, just because something new comes along doesn't mean everything that went before becomes obsolete and redundant.
Some of the things that can now be done with agentic AI may provide a better alternative than what was available previously. Some things will still be better, easier, or even faster with a dedicated solution and a non-text-based interface.

It's great to be excited by the new and the shiny, but that doesn't always mean the replacement of everything that went before.

---

I wonder what other abandoned side-projects I have that will turn out to be relevant in the future?

I hope that not all of my predictions about what's coming soon are as far out as the two I mentioned above.