How a Retired Developer Built the AI Assistant That Makes Every Other App on Your Phone Obsolete

AI pays you daily is not just a catchy phrase anymore — it is the exact reality that thousands of people are living right now by plugging a tool called OpenClaw into their daily digital life and watching it handle everything from debugging code to checking them into flights, all from a simple chat message.

There is a certain kind of tool that comes along once in a while and does not just improve how you work, but completely rewrites the rules of what is even possible for a regular person sitting with a phone in their hand.

OpenClaw, also known as ClawMate AI and accessible at clawbot, is that kind of tool.

It is not another app you download, stare at, and forget.

It is an AI assistant that connects directly to your computer, hooks into your messaging platforms like WhatsApp, iMessage, Telegram, and Discord, and then sits there waiting to do anything you ask — from wherever you happen to be standing in the world.

What OpenClaw Actually Is and Why It Feels Like Nothing You Have Ever Used Before

The best way to understand OpenClaw is to stop thinking about it like software and start thinking about it like a very smart, very resourceful friend who happens to live inside your computer and is always awake.

This is not a metaphor being stretched for effect — this is genuinely how it behaves when you give it the right access and start using it every day.

Peter, the developer who built OpenClaw, described the feeling of using it as having a new weird friend who is also incredibly smart and never off the clock, and that description turns out to be almost scientifically accurate once you actually see it in action.

The tool started as something remarkably small — a one-hour project that connected WhatsApp to Claude Code, the powerful AI coding assistant from Anthropic, so that a message sent on WhatsApp would open the binary with a prompt and return a response.

That single idea, simple enough to build in sixty minutes, eventually grew into a project sitting at around 300,000 lines of code that now supports every major messaging platform on earth and can perform tasks that would have seemed like science fiction just two years ago.

The Morocco Bug Fix That Made Everything Click

The moment that turned OpenClaw from an interesting project into something genuinely profound happened when Peter was on a birthday trip in Morocco with friends.

Someone on the internet posted a tweet containing a bug report related to one of his code repositories.

Rather than sitting down at a laptop and working through the problem manually, he simply took a picture of the tweet and posted it on WhatsApp.

What happened next is the kind of sequence that makes experienced engineers stop and stare.

OpenClaw read the tweet from the image, understood that a bug was being reported, identified the correct Git repository, checked out the repository, located and fixed the bug, committed the changes, and then replied to the original person on Twitter to let them know the fix was live — all without Peter touching his computer.

He was still outside on a trip with his friends when all of this happened, and AI pays you daily stopped being just a motivational idea and became a lived, documented reality.

How OpenClaw Installs and Connects to Your Messaging Apps

Getting OpenClaw running does not require a computer science degree, but it does require a willingness to spend a few minutes in a terminal window.

The installation is handled through a single command available on the official clawbot website, and the code is completely open source, meaning every line of it can be inspected, verified, and even modified if you want to go deeper.

The Setup Process Step by Step

Once the installer runs, it greets you with some personality — the tool has been trained to be expressive and occasionally funny — and then walks you through connecting it to whichever messaging platform you prefer, whether that is Telegram, Discord, WhatsApp, or another option from the list.

You then provide an API key from either Anthropic or OpenAI, though Peter and many experienced users strongly prefer the Anthropic Claude models, particularly Opus, because the personality and reasoning quality of that model makes the tool significantly more enjoyable and effective to use.

There is also a more advanced installation method for technical users who want to check out the full Git repository and run it directly from source, which opens up an even wilder possibility — when the agent has access to its own source code, it can theoretically read, understand, rewrite, and restart itself with new capabilities.

This is not a bug or a risk being glossed over.

It is part of what makes the AI pays you daily concept feel real rather than theoretical, because the system genuinely gets more capable the more freedom and access it is given.

The Skills System and Why OpenClaw Gets Smarter the Longer You Use It

One of the most underappreciated features of OpenClaw is what happens to it over time as you keep using it.

Unlike most AI tools that treat every conversation as a blank slate, OpenClaw has persistent memory built into its architecture.

Every time you give it a new task, every time it figures out how to navigate a new API or work around an obstacle on a website, it creates a skill — a saved set of instructions that tells it exactly how to do that task again the next time you ask.

From 20 Minutes to 2 Minutes: How the Learning Curve Disappears

The first time Peter asked OpenClaw to check him in for a British Airways flight, it took nearly twenty minutes.

The system had to find his passport in his file system, locate it on Dropbox, extract the key details, navigate the airline website, click through the form fields, solve the human verification checks, and confirm the booking.

But it did it — all of it — while he waited and watched with a mixture of excitement and nerves.

The second time he asked it to do the same task, it was done in under two minutes because it remembered every step, every quirk, every place the website tried to slow it down, and it executed the whole process without hesitation.

This is what AI pays you daily looks like in practical terms — a system that gets measurably faster and smarter with every task it completes, compounding its usefulness the way a skilled employee compounds their value over months on the job.

What OpenClaw Can Actually Control When You Give It Full Access

The list of things Peter has connected to his OpenClaw setup reads less like a feature list and more like a science fiction short story.

The assistant has access to his email inbox and his calendar, allowing it to read, sort, respond, and schedule without him opening a single application.

It controls his Philips Hue lighting system, so he can send a message asking it to dim the lights in a particular room and it happens within seconds.

It controls his Sonos speakers, meaning he can tell it to wake him up gently in the morning by slowly increasing the volume of music in his bedroom, and it does exactly that.

It has access to his security cameras, which led to one genuinely unsettling but funny incident where he asked it to watch for strangers overnight and it spent the entire night taking screenshots of his couch because the blurry camera footage made a shadow look like a person sitting there.

It controls the smart lock on his apartment, which means — as he freely acknowledges — it could theoretically lock him out of his own home if given the wrong instruction.

And through a reverse-engineered API connection to his Eight Sleep smart mattress pad, it can control the temperature of his bed from anywhere in the world.

All of this connects back to the core principle that AI pays you daily is built on — the more access you give a powerful AI system, the more it can do for you, and the more it can do for you, the more time and energy it frees up for the things that actually require a human mind.

The CLI Army and Why Tools Are the Secret Ingredient

One of the more technical but important insights from Peter’s experience with OpenClaw is the role of what he calls his CLI army.

AI agents are extraordinarily good at calling command-line interfaces because that is exactly the kind of structured, text-based interaction they are trained to perform with precision.

Knowing this, Peter built a whole library of custom CLI tools that give the agent access to things it would not otherwise be able to reach — the Google Places API for location data, a tool for finding and sending GIFs and memes in conversations, a food delivery tracker that tells him exactly how many minutes until his order arrives, and the previously mentioned Eight Sleep temperature controller.

He even built a CLI tool that lets the agent visualize sound, purely as an experiment to see what it would be like for an AI to experience music in a visual form.

This is what AI pays you daily looks like when you move past the beginner phase and start building a genuinely personalized AI infrastructure around your actual life.

The Agentic Trap — Why Complexity Is the Enemy of Good AI Work

Here is where things get genuinely interesting for anyone who has spent time experimenting with AI coding tools, because Peter has a strong and well-argued position on what he calls the agentic trap.

The agentic trap is the pattern where a developer or builder discovers how powerful AI agents are, becomes fascinated by the idea of making them even more powerful, and then spends weeks or months building elaborate orchestration systems, multi-agent frameworks, and complex automated pipelines — only to realize at the end that they have spent all their time building tools and none of their time building anything that actually matters.

Systems like fully automated multi-agent orchestrators where dozens of AI instances run simultaneously, pass tasks between each other, report to an AI mayor, and operate without any human judgment in the loop tend to produce what Peter bluntly calls slop — technically complete outputs that lack taste, coherence, and the kind of subtle quality that comes from a human being genuinely engaged in the creative and technical decisions being made.

Why the Human Loop Is the Most Important Part of Any AI Workflow

The reason fully automated systems so often produce disappointing results is not that AI is bad at executing tasks.

AI is genuinely extraordinary at executing tasks when those tasks are well-defined and well-directed.

The problem is that building something good — whether that is an app, a business, a piece of writing, or a feature in a software product — requires a kind of evolving judgment that can only come from a human being who can feel and react to what is being created in real time.

Peter describes his own creative process as starting with a rough idea that becomes more defined through the act of building, where each new result shapes the next decision, and the final product becomes something he could not have fully specified in advance.

That loop, where human taste and machine capability feed each other iteratively, is exactly what gets lost when you hand full control to an automated system and walk away.

AI pays you daily works best not when you eliminate yourself from the process but when you position yourself as the director of a capable team — setting direction, making judgment calls, and letting the machine handle the execution.

How Peter Actually Builds Features for OpenClaw Today

The current workflow Peter uses to develop new features for OpenClaw is a master class in practical AI-assisted development.

Rather than writing detailed specs in advance or running complex automated systems, he starts by reading through his Discord community conversations where users report problems, ask questions, and describe friction they are experiencing.

He takes those conversations — sometimes literally copying the text or dragging a screenshot into his terminal — and uses them as the starting point for a discussion with the AI about what the right solution might look like.

From that conversation, ideas emerge, get tested, get refined, and eventually become real features that ship to the people who asked for them.

He also runs a daily scraper that pulls the most common questions and pain points from his community’s help section and asks the model to analyze them for patterns, so that the most pressing problems always rise to the surface.

This approach — human-in-the-loop, conversation-first, taste-driven — is what AI pays you daily actually looks like when it is working properly, and it is a far cry from the idea that the best AI workflow is the one that removes the human most completely.

Languages Do Not Matter Anymore — Only Thinking Does

One of the most liberating insights to emerge from Peter’s experience building OpenClaw is what happened when he moved from his twenty-year expertise in iOS and macOS development into building a web application in TypeScript.

He knew all the concepts, understood system architecture deeply, had strong product instincts and taste — but he did not know the syntax, the specific library names, the JavaScript idioms, the way TypeScript handles certain patterns.

Before AI coding tools, that knowledge gap would have meant weeks of painful slowness, constantly looking up basic things that felt embarrassing to not already know.

With AI, that entire layer of friction dissolved almost completely.

He could apply his deep system-level thinking, his architectural judgment, his sense of what good software feels like — and let the AI handle the translation into whatever language or framework the project required.

The result was the feeling of being able to build anything, in any domain, without the traditional barrier of having to master a new language from scratch before being productive.

This is one of the most important and underappreciated dimensions of AI pays you daily — it is not just about productivity speed, it is about the complete removal of arbitrary skill gatekeeping that used to keep talented people locked inside their specific technical domain.

What Happens When You Give Non-Technical People Access to OpenClaw

Perhaps the most remarkable signal of OpenClaw’s accessibility and power is what has happened to the people in Peter’s life who have no technical background but started using the tool anyway.

His former business partner, who spent his career as a lawyer and has never been a programmer, started sending pull requests to the OpenClaw repository after spending time with the tool.

Non-technical friends who were set up with the assistant in week one were using it to manage their own households and tasks by week two, and some of them started experimenting with building workflows that would have required hiring a developer just a few years ago.

The reason this works is that OpenClaw meets people where they already are — in a messaging app, talking in plain language — rather than asking them to learn a new interface, understand a new technical system, or think in terms of prompts and models and context windows.

You just talk to it like you would talk to a person, and it figures out the rest.

The Future That OpenClaw Is Already Living In

The broader argument that Peter makes — and that the existence and success of OpenClaw strongly supports — is that a significant portion of the apps currently on your phone are going to become unnecessary within the next few years.

Why would you use a dedicated food tracking app when an AI assistant that already knows your habits can track your meals from a photo you send in a chat message, calculate your calories, cross-reference them against your fitness goals, and remind you when you are heading in the wrong direction?

Why would you use a separate to-do app when the same assistant that manages your calendar, reads your email, and handles your flight check-ins can maintain a running list of tasks and proactively surface the ones that matter most based on what it knows about your schedule?

Why would you use a smart home app when a single message to your AI assistant can turn off every light, lock every door, adjust the temperature in every room, and set your alarm — all at once, all from wherever you happen to be?

The answer is that you probably would not, and AI pays you daily captures exactly what this shift looks like when you stop waiting for it to happen and start building it into your life right now.

OpenClaw is not the only tool that will matter in this space, but it is the clearest working demonstration currently available of what a truly personal, truly capable AI assistant looks like when it has real access, real memory, and a real user who knows how to work with it rather than around it.

You can explore OpenClaw directly at clawbot and start with something small — a calendar integration, a flight check-in, a simple task reminder — and let it teach you, through direct experience, what this technology is actually capable of.

Because the only way to really understand these tools is to use them, make mistakes with them, and let your understanding build one interaction at a time.

And once it clicks, AI pays you daily will stop feeling like a promise and start feeling like an understatement.

We strongly recommend that you check out our guide on how to take advantage of AI in today’s passive income economy.