← Back to posts
Mar 27, 2026
Zero To Engineer: The AI Agent Your Phone Could Already Be Running
You have probably seen the word "agent" everywhere lately. Here is what it actually means and why it matters right now. (Where every Friday we discuss something new)
Hi friends,
You have probably seen the word agent showing up everywhere in tech conversations lately. AI agents. Agentic workflows. Autonomous agents. It gets thrown around a lot, and most of the coverage assumes you already know what it means.
I did not want to write that kind of post. So let me back up and explain this from the beginning, because what is happening right now is genuinely worth understanding, even if you are early in your journey.
-----------------------------------------------
First, what is an AI agent?
-----------------------------------------------
You already know what a regular AI tool does. You type something, it responds, then you go do something with the answer.
An agent is different. Instead of just answering you, it acts on your behalf. You give it a goal, and it figures out the steps, takes them, and comes back with a result. It can open files, run code, browse the internet, send emails, and interact with other software. It does not just tell you what to do. It does the thing.
Think of the difference between asking a friend for directions and handing your friend your keys and having them drive you there. Same destination. Very different relationship with the task.
What is OpenClaw and why is everyone talking about it?
OpenClaw is a free, open-source AI agent that went really viral in early 2026. It was built by a developer named Peter Steinberger and it hit 135,000 GitHub stars faster than almost any project in recent memory. For context, that is a lot of developers paying attention very quickly.
It runs on your own computer or phone. It connects to AI models you already use, like Claude or ChatGPT, and gives them the ability to take real actions. You talk to it through apps you already have, WhatsApp, Telegram, iMessage, Slack. And when you give it an instruction, it can read and write files on your computer, browse websites, send emails, control other apps, and run commands.
You are not chatting with it and then doing the work yourself. It does the work.
One developer set it up to build an app while he was grabbing coffee. Another used it to build a weekly meal planning system in Notion that saves his family an hour every week. Someone else had it running their coding agents overnight while they slept.
That is the shift. AI that works while you are not looking.
At NVIDIA’s GTC conference last week, Jensen Huang called OpenClaw the operating system for personal AI.
Not a tool you pick up and put down. An operating layer that runs in the background of your life.
-----------------------------------------------
But here is the part I want you to sit with
-----------------------------------------------
I am not telling you to go install this today. I want to be straightforward with you about why.
Within two weeks of going viral, OpenClaw had a serious security problem. Attackers got into its public skill marketplace and distributed malicious skills, small add-ons that looked legitimate but were designed to steal data or install harmful software. At one point, roughly twelve percent of the entire marketplace was compromised.
One of the maintainers of the project himself said publicly that if you cannot comfortably run a command line, this tool is too dangerous for you to use safely right now.
That is the creator of the thing telling you to be careful.
-----------------------------------------------
So why am I writing about it?
-----------------------------------------------
Because understanding what OpenClaw represents matters even if you never run it. The concept of a locally running AI agent that connects to your real tools and takes real actions on your behalf is not going away. It is going to get safer, more polished, and more accessible. And the developers who understand what these systems are and how they work are going to be better positioned to build with them, evaluate them, and make smart decisions about when to use them.
-----------------------------------------------
What this means for you as someone learning to build
-----------------------------------------------
Right now, most AI tools are assistants. They help you think, write, and code faster, but you are still the one executing.
What OpenClaw points at is a world where AI is not just assisting you but operating alongside you. Where you describe an outcome, and a system figures out the path. Where your tools talk to each other without you in the middle of every step.
That world is not fully here yet. But it is close enough that people are building serious infrastructure around it, fighting over its security implications, and integrating it into platforms with a billion users.
As someone building toward a career in software, knowing what an agent is, how it works at a basic level, and what problems it is designed to solve is becoming part of the baseline. You do not need to be an expert. You just need to not be confused when it comes up, and it is going to keep coming up.
We are in the early days of this. The best time to start paying attention is now.
Let’s Build It Beautifully,
Fab
You have probably seen the word agent showing up everywhere in tech conversations lately. AI agents. Agentic workflows. Autonomous agents. It gets thrown around a lot, and most of the coverage assumes you already know what it means.
I did not want to write that kind of post. So let me back up and explain this from the beginning, because what is happening right now is genuinely worth understanding, even if you are early in your journey.
-----------------------------------------------
First, what is an AI agent?
-----------------------------------------------
You already know what a regular AI tool does. You type something, it responds, then you go do something with the answer.
An agent is different. Instead of just answering you, it acts on your behalf. You give it a goal, and it figures out the steps, takes them, and comes back with a result. It can open files, run code, browse the internet, send emails, and interact with other software. It does not just tell you what to do. It does the thing.
Think of the difference between asking a friend for directions and handing your friend your keys and having them drive you there. Same destination. Very different relationship with the task.
What is OpenClaw and why is everyone talking about it?
OpenClaw is a free, open-source AI agent that went really viral in early 2026. It was built by a developer named Peter Steinberger and it hit 135,000 GitHub stars faster than almost any project in recent memory. For context, that is a lot of developers paying attention very quickly.
It runs on your own computer or phone. It connects to AI models you already use, like Claude or ChatGPT, and gives them the ability to take real actions. You talk to it through apps you already have, WhatsApp, Telegram, iMessage, Slack. And when you give it an instruction, it can read and write files on your computer, browse websites, send emails, control other apps, and run commands.
You are not chatting with it and then doing the work yourself. It does the work.
One developer set it up to build an app while he was grabbing coffee. Another used it to build a weekly meal planning system in Notion that saves his family an hour every week. Someone else had it running their coding agents overnight while they slept.
That is the shift. AI that works while you are not looking.
At NVIDIA’s GTC conference last week, Jensen Huang called OpenClaw the operating system for personal AI.
Not a tool you pick up and put down. An operating layer that runs in the background of your life.
-----------------------------------------------
But here is the part I want you to sit with
-----------------------------------------------
I am not telling you to go install this today. I want to be straightforward with you about why.
Within two weeks of going viral, OpenClaw had a serious security problem. Attackers got into its public skill marketplace and distributed malicious skills, small add-ons that looked legitimate but were designed to steal data or install harmful software. At one point, roughly twelve percent of the entire marketplace was compromised.
One of the maintainers of the project himself said publicly that if you cannot comfortably run a command line, this tool is too dangerous for you to use safely right now.
That is the creator of the thing telling you to be careful.
-----------------------------------------------
So why am I writing about it?
-----------------------------------------------
Because understanding what OpenClaw represents matters even if you never run it. The concept of a locally running AI agent that connects to your real tools and takes real actions on your behalf is not going away. It is going to get safer, more polished, and more accessible. And the developers who understand what these systems are and how they work are going to be better positioned to build with them, evaluate them, and make smart decisions about when to use them.
-----------------------------------------------
What this means for you as someone learning to build
-----------------------------------------------
Right now, most AI tools are assistants. They help you think, write, and code faster, but you are still the one executing.
What OpenClaw points at is a world where AI is not just assisting you but operating alongside you. Where you describe an outcome, and a system figures out the path. Where your tools talk to each other without you in the middle of every step.
That world is not fully here yet. But it is close enough that people are building serious infrastructure around it, fighting over its security implications, and integrating it into platforms with a billion users.
As someone building toward a career in software, knowing what an agent is, how it works at a basic level, and what problems it is designed to solve is becoming part of the baseline. You do not need to be an expert. You just need to not be confused when it comes up, and it is going to keep coming up.
We are in the early days of this. The best time to start paying attention is now.
Let’s Build It Beautifully,
Fab