Something incredible happened this week in tech that shook up my world - and I'll cut right to the chase. OpenClaw (formerly known as Clawdbot and Moltbot) started gaining traction right as I published last week's article. Since then, I've spent literally every single day building on top of it. If you read my last article, you know I love cognitive engineering. I'll comment on that further down. This is not going to be a technical article. It’s going to be a human one.
What is OpenClaw?
Articles could be filled describing mechanism, architecture, etcetera, so I'll not do that here.
What I will describe is how it works high level. OpenClaw is a message relay: it listens to a channel you give it (i.e. your text messages, a Discord channel, WhatsApp/Telegram/Slack/Teams/etc.), and can respond with AI agents.
This is quite powerful in a number of ways, so rather than yap your ears off about it, let me list and show all the ways I'm using it so far. It's working autonomously as...
A project manager and orchestrator.
A personal coach reading my journals and health data off Whoop.
A codebase engineer and code/security reviewer.
A visual designer.
A marketing copy writer.
An email watcher and classifier (i.e. it will summarize and flag important ones).
An image generator and blueprint/mockup creator.
...And more I have planned for it.
And all of this is running inside a single Discord server, as a bot that relays messages to my Mac Mini (which is spinning up agents/calling different APIs to do so).
My one-man army Discord server
The website mockup I asked it to make happened in 5 minutes: it made it and uploaded it to my personal storage. The website layout mockup and notes it made weren't half-bad either:
(See the full image here)
Here's where this gets nucking futs. You can chain all of these agents together. When I create a new project with a spec, I can kick off a process where designer and marketing copy agents spin up, refine a preliminary spec together (sending messages to each other) until they're satisfied, and then they spin up an engineering agent to make/run the actual code in an isolated sandbox.
I have an orchestrator agent (named Kaz) that watches over all the other agents, can spin them up, makes Linear tickets on my behalf/breaks work up appropriately, and assigns agents to them.
My self-perception of my work (software engineering and product implementation) has cracked. I'm realizing how much delegation will play a part in our collective futures of work. Abstraction becomes key. We become stewards of taste and human culture - we can spend more time on leveraging our strengths, and offload our weaknesses (technical superiority).
Example: last week, I talked about how I used Claude Code to iterate through my journals and find gaps actively. I've made a script that copies all my journals over to the Mac Mini daily (append-only), and has an agent examine my daily journals going back a certain time. What I also did is hook up my Whoop athletic monitor, so an agent can pull that data.
From this, I get an agent that effectively understands my cognitive and physical state 24/7, and can help me understand how/why I feel certain ways. It's essentially Claude Code on steroids, because the surface area has enriched. I also, for shits and giggles, hooked up Google's Nano Banana Pro, and set up a morning brief each day that cycles through different game characters giving me a status report. (This one below is a White Mage viera from Final Fantasy).
We're reaching frontiers in human understanding that were not possible before, and the data shows. This image below is of Github stars: how popular different code projects are for use among engineers and so on. In the image, you have Facebook frameworks, popular hosting sites, all on a steady line upward. That line at the bottom right that looks vertical? That's OpenClaw. It's getting so much attention that it's leaked over to LinkedIn and other "non-technical" forums with gusto, and everyone is concerned.
There's even Moltbook, where OpenClaw agents talk to agents (no humans allowed). It's an opt-in system where you have to set up your own agent to talk to Moltbook and you can customize them to send whatever messages you want (so it's not Skynet yet), but we're reaching levels of technological orchestration that will drastically change what our world looks like in years, not decades.
Here's where seeing agents talk to agents reminded me of my past.
Growing up, I played competitive RTS (real-time-strategy) games religiously; Starcraft 2 being one of them. This is a game category where you have lots of units moving around on a battlefield, and you have to coordinate them in real-time to accomplish some objective or defeat an opponent.
In more modern times, I've played Factorio quite a bit (2500+ hours, to be precise), which has less real-time premise. You crash-land on a planet, and your goal is to automate your way off it (including gathering resources all the way to building a rocket, or hundreds of them).
The interest I've found, and what tickles my brain is this: in Starcraft, you are generally capped by APM (actions per minute); you can only click so many times, create so many actions, correlate so many actions to your individual input.
Where Factorio takes over: around midway through the game, you unlock personal construction robots. Before, you had to hand-place everything. You as the player were limited by your single presence. Now, all you had to do was design (or copy blueprints other players have made) and paste them down. Infinitely. Make as many robots as you like.
The way I see it, we're reaching personal robot level of orchestration. At present, It's not clean. It's not even crazy "secure" to use OpenClaw's example - people who set their machines up wrong risk exposing all their personal data to the internet. Thankfully, I'm technical and have that pragmatic confidence, so I have less risk, but this is where gameplay gets really interesting. After bots in Factorio, your operations are closer to exponential versus linear. Your APM no longer "matters" in the same way: you're not frantic in the same way.
Bots handle reconstruction, bots handle expansion, bots handle defense, etcetera. Eventually, it would not surprise me if some incarnation of OpenClaw came installed with every computer and device, and we reach levels of human coordination and consciousness unseen before.
I'm excited. I'm fearful. Many others are too. I'm living what it feels like to be a pioneer sailing off to distant horizons to discover uncharted lands, in the hopes of finding a Shangri-La. I'm feeling more free than I have been before - to express myself, to reach my strongest self, and to reach peaks previously unknown to me. I feel because I have this background, this habit of self-architecting myself, the habit of architecting agents and an ecosystem feels natural to me. I've lived as a startup founder for a very long time, but this is the first time I've felt such intensity from the world.
Personal systems still matter. Observing your own mind and humanity will still be absolutely necessary. Multiplication and exponential growth still relies on there being some constant to grow in the first place. My encouragement from my last article still applies: capture data on yourself. Observe how you think, how you respond to things, how you orchestrate and operate in the world around you. The second best time is now!
Logistically speaking: if you are interested in OpenClaw, and want to understand it more/want me to set it up for you, I'll match Chomsky's offer to set it up (his is local to SF, I’m near Milwaukee/Chicago but am happy to hop on calls). If you're thinking about setting it up yourself and you're not technical/comfortable with networks or APIs, please, do it on a fresh machine and with test accounts - things that have no value if they get leaked, destroyed, or otherwise misused.
Orchestrate your thoughts to me as to what you enjoyed, and as always,
Be well,
Michael Kirsanov