Agents are the new apps

In my previous article, I talked about how human–computer interaction has evolved. As the web itself went through its own transformation, the way humans communicate with computers has also followed a similar path.

If you haven’t read that piece yet, you can find it below.

I was planning to publish this follow-up right after finishing that article, but the past few weeks have been unusually intense, and I couldn’t carve out the time to sit down and write.

Then, in the meantime, Sam Altman and his team announced something that perfectly echoed what had been sitting in my mind, a new phase that makes the future of human–computer communication even clearer.

A version of ChatGPT where we can talk directly to applications.

This, to me, marks a fundamental shift: applications are no longer the front-end layer where users directly interact. Instead, they’re becoming background capabilities, skills that agents can use. We’ll simply connect our credentials, give our agents permission to access them, and let the agents handle everything on our behalf.

At this stage, agents already have certain abilities, they can use a terminal, operate a virtual computer, generate content, and build workflows.

Now, another layer has been added: the ability to use applications.

That means soon, the apps we use every day might no longer live on our home screens. Instead, they’ll exist as callable skills, orchestrated by whichever LLM-powered agent we use. You can think of this agent as a supercharged version of Siri — not something built for one specific action, but an orchestrator capable of managing all of our applications through a single, intelligent interface.

And what about operating systems?

When users start performing their tasks through an LLM that interacts with apps on their behalf, operating systems will inevitably evolve too. This shift won’t happen overnight — it’ll take time, and it’ll feel strange at first. But in the long run, apps as we know them today will become nothing more than APIs — no visible icons, no windows, no screens. The visual layer between humans and software will dissolve.

Of course, this transition won’t be simple. It challenges years of business models, interface conventions, and design patterns.

How will monetization work when the “user interface” disappears?

Can every product experience truly adapt to a conversational form?

Will users trust agents to make decisions, take actions, and represent them across tools and platforms?

These are difficult questions — but they also define the frontier we’re stepping into.

I believe the next era of computing won’t be about building apps anymore, but about building agents that use apps for us.

The interface won’t be a screen; it will be a conversation.

And sooner than we think, operating systems themselves will become conversational layers, environments where we simply describe what we want, and the agents handle the rest.