For years, no code promised to democratize web creation. It kept part of that promise — and broke another, deeper one. Now, AI agents are rewriting the rules in ways no one quite anticipated.
A Terminal in a Travel Agency
A few years ago, I was training a team on WordPress at a travel agency. When I arrived, I noticed something unexpected on the agents’ screens: a black background, green text, and strings of characters I didn’t recognize.
It was Amadeus — the global reservation system used by tens of thousands of agencies worldwide. No graphical interface. No buttons. No dropdown menus. Just commands to memorize, a syntax to master, and in return: remarkable power.
In a few keystrokes, an agent could query dozens of airlines in real time, compare fares, book a flight, issue a ticket — all in seconds.
What I saw that day was a system designed for direct dialogue between human and machine. No intermediate layer. No simplification that constrains. Just a shared protocol, used globally, in the hands of those who had learned it.
Graphical Interfaces: The Progress That Closes Doors
Graphical interfaces changed all of that. And it was necessary — at first.
Point and click instead of type and memorize: that was an accessibility revolution. Millions of people could use computers without ever learning a single command. That’s real progress.
But something was lost in the process.
By making systems more accessible, graphical interfaces also made them more closed. You interact with what the designer chose to expose — and nothing else. Menus replace commands, buttons replace scripts, and gradually, users lose the ability to talk directly to the system.
They no longer pilot. They navigate a pre-defined path.
No Code: The Peak of Closure
The no code movement pushed this logic to its extreme.
The original idea is compelling: let anyone build apps, websites, automations — without writing a single line of code. And on paper, it works. Tools like Webflow, Bubble, Zapier or Notion have enabled non-developers to build useful things, quickly.
But no code has one critical blind spot: closure by design.
Every no code tool is a box. An ergonomic, well-designed, often elegant box — but a box. You can do everything the designer anticipated. And nothing else. When you hit the limit, there’s no escape hatch. You don’t have access to the engine. You either accept the constraint, or start over in a different tool.
The paradox of no code is that in trying to empower non-developers, it created a new form of dependency — more comfortable, but just as real.
What’s Changing With AI Agents
Over the past few months, something significant has been happening.
Tools like Claude Code, OpenAI’s Codex, or Cursor are not just writing assistants. They are agents capable of reading a project, understanding its architecture, writing functional code, running it, fixing errors — and starting again.
That fundamentally changes the no code question.
If an agent can translate a natural-language intention into executable code, then the barrier that justified no code — “I don’t know how to code” — collapses. You no longer need to know Python or Bash to run a custom script. You need to know what you want to do. And be able to describe it.
That’s exactly what the Amadeus terminal offered — but with a radically different level of accessibility.
We’re going back to the terminal. An augmented terminal. A terminal you can talk to.
What I’ve Done Concretely
Over the past few months, I’ve replaced about twenty services and apps with Bash scripts built with Claude Code.
Scripts to:
- Convert audio files into structured transcriptions using Whisper.
- Generate automated reports from WordPress data via WP-CLI.
- Sync content between development and production environments.
- Automate exports and backups based on precise business rules.
Each script is configured exactly for my needs. No subscription. No limits imposed by a vendor. No interface hiding what’s actually happening.
The result: I work faster, I understand what I’m doing, and I’m no longer dependent on someone else’s product roadmap.
The Future Is Neither Code Nor No Code
What’s emerging is something different from both previous models.
Not raw code, reserved for developers who know the syntax.
Not closed no code, accessible to all but constraining by nature.
But an open space where intention is enough — as long as you know how to talk to the system.
This isn’t magic. You still need to understand what you want to build, have a logical approach to problems, and be able to evaluate a solution. AI agents amplify capability, they don’t replace reasoning.
But for those willing to relearn how to talk to machines — not by memorizing syntax, but by building clear intentions — the possibilities are as open as they haven’t been since the terminals of the 1980s.
Amadeus understood this before anyone else.
Where Does WordPress Fit In?
WordPress is particularly well positioned in this new landscape.
It’s a structured system — database, REST API, custom post types, taxonomies — that can be queried, driven, and enriched by AI agents. On several recent projects, I’ve implemented WordPress as an MCP node (Model Context Protocol): Claude can read, create, and manage content directly via structured prompts.
This is no longer “WordPress with a bit of AI.” It’s WordPress as the control layer of an intelligent system.
The terminal is back. And it speaks WordPress.