AI Coding: A Meta-Capability
Yesterday, while introducing our product to a friend, I suddenly realized something.
Underlying technology is evolving rapidly—to what extent? It's breathtaking. Yet the general public—even many white-collar professionals—are adopting these tools far more slowly than you'd imagine.
Here's a recent example. OpenClaw went viral, and everyone was captivated by the technical demonstration. Someone asked me: "Are you using Claude Code? Is that for helping people write code?"
I said yes.
Their first reaction was almost identical: "That sounds complicated. I probably won't need it in my lifetime."
It's Not What You Think
I demonstrated in the most intuitive way possible: five minutes to set up, you just need an account to get started, and all you need is the ability to speak.
They stared at my terminal for a moment, then fired off a series of questions. Is installation complicated? Will it be difficult? What problems can it solve?
I said, installation is just one line of code, five minutes. An account costs just a few dozen dollars per month to get started.
They were stunned. "Is that how it is? I thought something this powerful would have a complex installation process."
Not at all.
What's even more surreal is OpenClaw itself. Online, you can find 15-step, 20-step installation guides that look headache-inducing. Many people complain: "The lobster works great, but installation is too difficult." On Xianyu (a second-hand marketplace), services offering door-to-door installation even appeared for 500 RMB.
I demonstrated right then and there. Once Claude Code is installed on your computer, you only need to say one sentence—"Help me install OpenClaw on this computer."
It will find out what OpenClaw is, where to download it, and how to deploy it on its own. You tell it the Feishu (Lark) App ID, and it will help you debug and test until it's running. You go grab a cup of coffee, and when you return, the deployment is complete.
One line of code, five minutes. That simple.
A Meta-Capability
Someone asked me: "Does your software have out-of-the-box applications? Like text-to-image?"
I didn't answer directly. I was thinking about something else.
If we enable everyone to master an efficient AI coding tool—a coding agent—then whatever you want, just ask it to install it for you.
Text-to-image? You say: "I want a text-to-image application, download the model and deploy it locally so I can generate images with a single sentence." It might take an hour or two to download the model, deploy it, and get it running.
This capability has already transcended the concept of an "application."
It's helping every ordinary person—at an extremely low barrier to entry—truly take control of their computer. By "control," I mean to a degree that exceeds 90% of engineers and senior engineers. And your computer is connected to the internet.
So the previous mindset of "I need to install an out-of-the-box application" is actually unnecessary. Just tell it what you need. It will do it for you.
What Can It Do?
Honestly, at this point, if I had to use a computer without such software, I wouldn't know how to use it.
It can unlock the 90% of your computer's capabilities that you normally don't use.
You have 20 Excel files to process? Put them in a folder, toss it to the agent, and tell it how you need them analyzed. It will invoke Python; if the plugin isn't there, it will download it itself; if it encounters problems, it will solve them itself.
Want to implement some interesting network functionality? It can help with that too.
As long as the model is strong enough and fast enough, these tasks that originally required specialized skills become within reach.
What About the Future?
In the future, people will become increasingly lazy when using computers, unwilling to do things manually. I rarely operate a computer myself anymore—at most, I open a few Claude Code windows and switch between them to have them do different tasks.
The way we work will change.
Previously: Human → Tool → Computer.
In the future, roughly: Human → Agent → Computer.
What's the point of building a complex interface? It's a burden for AI, and humans don't look at it either. Just give instructions to the agent and let it execute. Done.
Software infrastructure might change. I'm not sure what it will become, but this is the general direction.
The Difference Between Claude Code and OpenClaw
Many people ask me this question.
Strictly speaking, OpenClaw wraps a layer around Claude Code's capabilities. Some people call this wrapper the "lobster layer"—I think that makes sense.
What Claude Code does: You give it a vague instruction, and it executes end-to-end, basically meeting expectations during the process without requiring your intervention. It can run for dozens of minutes or even several hours, finally delivering a result.
What OpenClaw does: It adds another layer on top—distributing instructions, managing results, forming long-term memory, acting like a project manager or supervisor. It can interact with you at a higher level, becoming your personal assistant.
But all of this depends on one premise: the underlying coding agent must execute each task beautifully, without you having to watch over it.
This is also why OpenClaw connected to different models is an entirely different species.
Currently, there aren't many models that can truly "execute well on a vague instruction." I've seen only Claude's Opus, and possibly Codex's latest model, capable of this. Versions connected to other models feel like toys—there's no guarantee of execution each time, making the coordination layer above even more difficult.
The reason OpenClaw amazed everyone is that it was connected to Opus. But Opus's coding plan is no longer available now—too expensive.
This is the current pain point. I don't know when it will be resolved.
Originally published at https://guanjiawei.ai/en/blog/coding-agent-meta-ability