I tried OpenClaw, so you don't have to...

I finally decided to give OpenClaw a try. I was just waiting for it to get a stable name 😀

If you’re in the X bubble, you probably got tired of OpenClaw posts by now. But if you’re someone who actually has a personal life and doesn’t scroll X every 5 minutes, here’s the deal.

OpenClaw is an open-source personal AI assistant that you run on your own machine. You connect it to Telegram, WhatsApp, Slack, or whatever you use, and it becomes your always-on AI agent. It can browse the web, schedule tasks, pull data from APIs, and basically do things on your behalf.

Now, you might be wondering: how is this different from just chatting with Claude or ChatGPT?

Here’s the thing. When you chat with Claude, you get answers. When you set up OpenClaw, you get an assistant that lives on your machine, has access to your data, and can actually take action. It’s the difference between asking a friend for advice and hiring someone who shows up every morning and gets stuff done.

That’s a big difference.

My experience

Yesterday, I decided to give it a go. I spun up a VPS server, set it up, and shortly after that I remembered that I have my old MacBook Pro from 2019 sitting in a drawer. It’s an Intel-based machine, nothing fancy. But it’s more than enough.

Just a side note here. You don’t need a powerful machine to run this. In most cases, the heavy lifting is happening remotely. You’re basically invoking LLMs like Anthropic, OpenAI, etc. The machine is just the middleman.

So before you go and buy a Mac Mini, check if you have an old laptop lying around. If not, grab a cheap VPS for $5 a month. That’s enough to mess around with it.

I factory reset the MacBook, installed OpenClaw, and made sure it will never go to sleep, even if I close the lid. Now it’s an always-on machine.

I set up TeamViewer so I can remotely access and manage it from my main machine. Then I connected it to Telegram.

I gave it a name: Danchik. Same as my dog. I was just pissed that he’s not paying any rent. Now, at least his digital version will contribute to the household, while the real one is just chilling on the couch.

At this moment, I was like… ok. I now have Claude, but in Telegram. Cool. Now what?

Then I couldn’t sleep

Yesterday, I was not able to fall asleep. I was lying there, thinking about all the possibilities you could achieve with this if you just connect enough data to it.

I believe in a couple of years months, a system like this, one that has access to your data and is smart enough to work together with you, will replace most software. Why would you need a separate gym logger & a food logger, if you can have this built and managed by your agent, so it has the full picture of your food & fitness. And you could just record a short voice note, describing you workout. Then, it understands it, and logs everything for you. No manual entries.

This is just an example to show you what I mean.

But to understand it yourself, just give it a try, and actually force yourself to delegate some tasks and decisions to it.

Anyway, back to some practical stuff.

Giving Danchik some skills

Today, I woke up and decided to give Digital Danchik some actual skills.

First thing first, I gave it access to my RevenueCat data. This will help him understand the revenue data about my apps, which will lead to better advice and better workflows for growing my portfolio.

This was surprisingly straightforward. I set everything up just by chatting with it. The only thing I had to do manually was create read-only API keys and give them to him. Then, he figured out everything else.

I created a new Telegram topic, and I asked the agent to send me reports each Monday morning. That’s it. He scheduled a cron job, and I expect a report next Monday.

Still, this might seem trivial. But this is just the beginning.

Where this gets interesting

Imagine this. I ask Danchik to help me improve the LTV of one of my apps. It checks the revenue data across all my apps. Looks at the ones with good LTV. Analyzes what leads to increased LTV. Which experiments, which commits, which changes made the difference. Then it recommends a plan.

And not just recommends, but it goes ahead, implements the changes, and creates a PR that I can just review and merge.

Is that possible today?

Yes!

In the following weeks, I’m going to push the boundaries and show you what’s actually possible and what’s not.

A couple of tips if you try it

Back up your OpenClaw workspace to GitHub. This makes it versioned. If you mess something up, you can go back. It also allows you to move it to a different machine if needed.

Also, read more about security. This is currently the wild west. A lot is possible, but at the same time, the security can’t keep up, and there are real risks. Don’t expose your instance to the public internet. Be careful with the permissions you give it.

Why am I doing this?

Things are changing fast. I want to stay on top of the new stuff. To stay relevant.

We are going through a revolution bigger than the internet. That comes with challenges and opportunities. That’s why, even if I might not see a direct benefit from a new AI tool right away, I will give it a try and play with it like a child.

Just by pure exposure, you will have a better understanding of the space, will see where things are moving, and will spot the opportunities before everyone else.

Otherwise, you will wake up in 6-12 months wondering wtf just happened, and will miss the good old days when we used to write useEffects by hand.

Don’t be that guy.

Don’t be the guy who refused to adapt and stuck to WordPress because “it just works.”

That's it for now!

Cheers, Vadim

PS. Hit reply and tell me: how are you using AI nowadays? I’m curious to learn new flows and tools.

Tools I use to build & ship apps fast:

• ​Screensdesign generates amazing onboarding flows and UIs with their AI model trained on thousands of apps
• ​Astro for ASO keyword research. All my app ideas start here.

*Some of the links are affiliate links, but I will only shout out tools I actually use in my flow.