Drag-and-Drop Is Over. Just Tell AI What You Want

In the early 2010s, “no-code” and “low-code” builders exploded. Zapier, Webflow, Bubble and others gave us a friendly abstraction layer so anyone could spin up automations, landing pages, and even fairly sophisticated web apps without touching a text editor. Millions of people and businesses built on top of these platforms, and a whole generation of “builders” emerged.

Then everything changed.

Starting with ChatGPT in late 2022, and cemented by tools like Lovable a few years later, anyone could just say what they wanted and watch a working app appear minutes later. The kicker: these apps were written directly in code. No new proprietary language to learn, no platform to master, no gatekeeper to pay. LLMs made it possible for the average person to create real, portable software by asking in plain English.

The Abstraction Trap

This is exactly where things got rough for Zapier, Webflow, Bubble, and the rest. The abstraction layer that once made them magical became their biggest liability.

• Over time, the once-simple drag-and-drop UI turned into a coding language in disguise; a maze of boxes, wires, and settings that takes serious effort to wield for anything beyond a toy app.
• Worse, LLMs don’t naturally “think” in these UIs. They think in code, not in some complicated, nested JSON structure required to manipulate legacy builders.

So consumers now face a very simple choice: Learn a bloated, 10-year-old piece of software to assemble an app… or type what they want and get actual code.

“We’ll Add AI!” …But AI Wants Code

To their credit, most of these companies see what’s happening and are racing to bolt on AI assistants. The problem: they’re running uphill. The major AI labs like OpenAI, Google, and others are pouring tens of billions into models that excel at writing code, not at puppeteering bespoke, proprietary UI schemas.

Time will tell, but the writing’s on the wall: these legacy UIs are on borrowed time—no matter how they rebrand or which cutesy mascot they slap on their “AI assistant.”

If AI Can Code, Why Are We Still Dragging Tiles?

Why should anyone drag tiles around a canvas and connect them with cryptic lines? Why force users to pre-define elaborate workflows before they can do anything useful?

Lumetric’s Take

At Lumetric, our goal is simple: give everyone the power to create agents that do what you ask, via the channels you already use. No new platform to learn. No new language to memorize. No heavy abstraction to wrestle.

Just tell your agent what you need (over text or email) and it gets done.

• No new platforms. Use the tools and channels you already live in.
• No new languages. Speak in plain English.
• No abstractions. Skip the diagram gymnastics and let AI generate the logic.

Drag-and-drop was a great bridge. But bridges are for crossing. In a world where AI can code better and faster than any human, the path forward is obvious: ask for what you want and just get it.