Intro

In a saturated market of AI projects, products and services, a new, yet very inspiring story has crawled its way to AI enthusiasts, engineers and the like.

If you’re a normal person trying to figure out how to use them AI agents and see what the hype is all about, look no further; you’ve come to the right place - OpenClaw(opens in new tab), the viral AI agent that took the world by storm thanks to its configuration simplicity.

It’s a nice little app that works on your machine, uses the agent subscriptions or open-source models of your choice and you can control it from anywhere in the world with your favourite chat app.

The open-source agent that clicked

I heard about OpenClaw for the first time when it was still called Clawd, but it has since changed its name, following a lawsuit inquiry by Anthropic over name similarity with Claude. In all honesty, I don’t blame them. With the quick rise in popularity of the AI agent orchestrator, confusion is to be expected.

I heard about it when a friend of mine told me “Have you seen this? I’ve been obsessed for days, I can’t stop using it”. This is how most people working with AI feel nowadays - overwhelmed at all times. Even the author of the nifty tool, Peter Steinberger, admitted that “vibe coding took over his life at one point”.

This obsession comes at the cost of brutal confusion in the midst of bottomless rabbit holes. This sudden empowerment that we got from using AI agents has quickly become a burden. The ability to produce any kind of content in the form of tools, documents, media, etc. gives this illusion of productivity, but what happens most of the time is that we create clutter: brainless prototypes, one after the other. We hit complexity and start wondering how to make it work, and the development process becomes slow and dreadful. If that’s the case, why not fully automate it?

Even Peter himself said that he’s no longer reading the code that he’s shipping. When the social network for bots got hacked(opens in new tab), he just said “software is prone to be vulnerable, no matter how you build it” (or something like that - I can’t find the actual quote now).

Overnight hysteria

It seems that the simplicity of the “bot for the people” has been more than lucrative. People all around the globe are buying Mac minis for the ability to use the bot (it works nicely in a Linux/Unix environment). Why a separate machine? Because it gets full control of it, so one has to be careful. In that regard, numerous memes have appeared. Here’s one of my favourites:

So in times of RAM shortages manufactured by premature sale expectations, people are now hoarding the Mac minis to run bots. I found one place reporting that a single user ordered 40. Imagine the scale!

Fact is, most people cannot really figure out how to organise agents to do the shit that they want and suddenly there’s one “agent for the people” that is open-source, free to get and everybody’s chomping on it.

I can’t deny that I myself feel tempted to install OpenClaw, but I haven’t done so yet. I just don’t feel OK with exposing my entire environment and secrets just like that. I’d make my own bot and control it, but borrowing thy neighbour’s still feels sketchy to me (on top of the hacks and new Internet trivia).

The other news on the agentic horizon

Everybody’s trying to find a way to optimise their workflows, to use agents to automate trivial, boring work and then take it to the next level where in the process we improve quality and processes and gain time.

At the beginning of January 2026, Anthropic introduced Claude Cowork(opens in new tab) - the Claude app for non-developers. Give it a folder of Excel files, images and docs and it will sort out a presentation for you in a matter of minutes. Organise my emails - easy-peasy. Plan my trip - done.

Then a few days ago, on February 2nd, 2026, OpenAI dropped their Codex app(opens in new tab). Similar approach to Claude Cowork, but using the OpenAI models.

This is happening everywhere. You had your prompts, now let’s convert them into workflows. And here’s why OpenClaw is kicking both OpenAI and Anthropic in the shins: it gives something they don’t - multi-agent capabilities and consensus across agents.

Conclusion

Adoption of agentic workflows by less technical users is the new now. I’m still able to meet those innocent people that live in a world without AI, oblivious of the missed opportunities, happily heading home after office hours and enjoying life as a normal human being.

For the rest of us that want to leverage this new technology, we’ll keep struggling to find ways to put the chaos in order. I’m personally still reluctant to give agents all my own data, but will definitely explore running OpenClaw in a controlled environment to see how it performs in orchestration of cross-agent software development.