An interesting thing is happening in tech right now, is that after decades of building complex visual interfaces, we are going back to text. Most AI interactions are chat first and text-based agents are everywhere. It feels like we're witnessing a renaissance of text-first computing, similar to the early days of the internet.
But why so? Great design teams and places with talent and financial resources, can't come up with something better than chat? Most AI interactions are prompted by us and blocks of text as answers. The easy and the most obvious reason is that LLM are “language” models at the end of the day so it makes sense to keep the interaction text first. But upon further research it feels like there are multiple things at play here.
Before I get into that let’s take a step back and see how we got here.
When the internet began, everything was text. On January 1, 1983, the internet's official birthday was marked by the adoption of TCP/IP, a text-based protocol that allowed different kinds of computers on different networks to talk to each other. Web pages were simple documents you could read. Email was plain text. If you wanted to tell a computer what to do, you typed commands.
This wasn't because we lacked better options. Text dominated for practical reasons. Bandwidth was limited, so sending images and graphics was slow and expensive. Processing power was scarce, making visual interfaces sluggish. Most importantly, the internet grew out of academic and technical communities where people were comfortable with text commands. Text had superpowers we've since forgotten. It was readable by both humans and machines. You could copy it, edit it, and share it easily. If something broke, you could see exactly what was happening because everything was written in plain language. The early web let you view source on any page and understand how it worked.
A key point ,text also matched how humanity was exploring this new technology. When something is truly new, you experiment with the simplest, most direct methods before adding complexity.
As personal computers became popular in homes and small businesses, we decided text was too cumbersome for regular people. The widespread adoption of PCs created demand for interfaces that didn't require technical knowledge. So we built graphical interfaces with windows, icons, buttons, and menus. These made computers easier for beginners, but they came with hidden costs.
Visual interfaces created walls between different programs. Your data lived in separate apps that couldn't talk to each other easily. Moving information around meant copying and pasting, exporting and importing, or learning entirely new interfaces for each tool. We also lost directness. Instead of telling the computer what you wanted, you had to navigate through menus and click through options. One had to learn each app's unique visual language. We won accessibility, but we lost simplicity and power.
Now coming back to my original point. Why have we collectively returned to text as an interface? First, AI finally understands natural language reliably. Earlier attempts at text interfaces failed because computers couldn't interpret what humans really meant. The breakthrough came with large language models trained on massive datasets. These systems can handle the full complexity of human communication, including nuance, context, and ambiguity.
Second, we've hit the limits of visual complexity. Modern software has become so feature-rich that finding anything requires searching through endless menus. Text interfaces let you skip directly to what you need by describing it.
Third, text can handle complexity that visual interfaces struggle with. You can describe intricate workflows, conditional logic, and multi-step processes in a few sentences. Building the same thing with buttons and menus would require dozens of screens. I believe we're at a similar turning point to when the internet was first discovered. We're still early in understanding how to use this technology effectively at scale. That's why text-based interfaces are returning. They represent first principles thinking.
Text-based systems dont hide what is happening. Every prompt, response, or command is visible and legible as plain text. Users, especially those new to computing with AI, can see exactly what the agent did, and can copy, modify, or retry commands easily reinforcing trust and learning. This closure in the interaction loop (you type → system responds) helps users understand cause and effect, fast. An effective UX tool for familiarity.
For people to adopt powerful new technology, you can't change everything at once. Too drastic a shift scares people away. Innovation needs familiarity as a foundation. Text gives us that familiarity because it's how we naturally communicate. Beneath all the progress, the chips, the models, and the interfaces, technology still has the same job it did at the start: to understand us. Text remains the most universal way we share ideas, teach, and tell stories. The return of text-first interfaces in AI is not a step backwards. It is a step towards meeting technology on our terms.
Visual computing will not disappear (as a visual designer i hope it doesn't), and text will not dominate forever. The two have always traded the spotlight as technology evolves. We are in a moment where text feels natural again, direct, flexible, and deeply human, while AI learns to interpret it with unprecedented nuance. This is not the end state. It is a bridge, and like all bridges, its value is in where it can take us next.
STAY
IN
THE
LOOP
NO SPAM. OCCASIONAL CERTIFIED GOOD STUFF.
PLS
COME
SAY
HI :)
