Most of us have been using AI Chatbots for all sorts of reasons for quite some years now. Some use it to automate work, some as a therapist, some as a friend (a sycophantic lackey), some as an intellectual sparring partner, and some even create clones of themselves (at least seriously try). I am guilty of all of the above. Spending 8 months as a Product Manager at an AI firm, where I spent so many hours shouting at Claude every day, didn't help much. At this point, I am addicted.
So what does an addict do? Find something with a stronger hit.
That was OpenClaw. It was all the rage in the AI-verse in the last 2 months. Many of you would have come across Moltbook memes as well. At its core, it's a simple product – it basically gives LLMs persistent memory, a toolkit, and the ability to run autonomously on your device. Everything local, everything open source, and everything configurable by you. The best model connection (as of when I wrote) was Opus 4.6, although Anthropic banned regular users from using that subscription with API access, so I am running with two Chinese models: Kimi K2.5 for demanding tasks and MiniMax 2.5 for regular cases. Chinese models are surprisingly good, but the best part is that they won't bankrupt you in 2 weeks, given the current state of Claude API pricing.
Persistent memory is the biggest game-changer for me. Most chatbots rationalize memory – they can't load everything into every conversation because of the token consumptions, so they summarize and compress, and eventually lose the texture of who you are. OpenClaw stores everything in plain-text files (.md) – and if you configure it correctly, everything will be structured in an Obsidian vault, which is fully browsable. It won't forget anything, and can retrieve whatever you want whenever you want. The toolkit gives it external access – web search, email, WhatsApp, etc. Autonomous running (self-scheduling cron jobs) means it can act on its own schedule and do things for you, rather than just waiting for you to prompt it.
These don't seem like much, but they are genuinely powerful. So powerful that Reddit is full of hilarious, scary stories. I think my OpenClaw knows enough to put me behind bars??? I am not sure, but it could do some major damage.
But it was so much fun setting up. I need to do a lot of scripting on the terminal. And it also behaves just like another Chatbot, unless you set up autonomous actions, a proper memory system, and give it some tools to act. Configuration is everything.
One of the first things we do is give our OpenClaw a name. Such an open-ended question. A question that is bound to send people into an existential crisis. Naming something is so hard, and I was building my OpenClaw to be a personal assistant that keeps me in check – like Alfred in Batman.
So I wanted to give it a meaningful name. Edison felt too successful. Gump too innocent, and slightly reta****. Taleb too combative and intelllectual. Darwin is too detached - I am a natural selection person, not a Darwin person. Rocky went to the gym agent immediately - he earned it without a second thought. Roark was what remained. And also, the only name that felt less like an aspiration and more like an obligation. So, Roark it was.
For those who don't know, Roark (Howard Roark) is the name of the lead character of Ayn Rand's The Fountainhead. Rand is a polarizing writer and I understand why. But read the book before deciding how you feel about it.
I came across The Fountainhead in 2021, and reading it was the most relieving and validating experience of my life – and like any impactful book for anyone, it didn't teach me anything new, but it put precise language around something I have felt several times strongly before, but couldn't say out loud. It also probably has a lot to do with where I was in my life.
The book centers on 2 architects. Howard Roark builds only what he believes in, and takes only commissions that don't require him to compromise his vision, and watches his career almost destroyed because of it. Peter Keating is brilliant and charming, and he builds exactly what clients ask him to and what the public expects a top architect to build. He becomes the biggest name in the city, wins all the awards, and gets every top commission, but still harbors a secret resentment towards Roark, who is practically a nobody, once worked under him, and has no major projects. Towards the last third of the book, he has everything, but feels like a nobody, driving him mad. The book is not a contrast between good and evil. Not even close, and that's what makes the book so devastating. Keating just made a series of reasonable choices, each small enough to justify, and all, until when he looked back, nothing was ever his own.
The Keating arc is the most devastating in the entire book. Towards the end of the book, in a long (really long) conversation with Roark, Keating, stripped of everything, confesses his inner desire has always been to become an artist, shows Roark a canvas, and asks him, "Is it too late for me?" I have shed a tear or two reading that.
A question that wouldn't stop haunting me. I am extremely close to being 27, and I am afraid.
It is very easy to read this book and think the problem Rand describes is specific to Architecture or design, or to any field with an obvious and personal aesthetic vision that you choose to betray. It is not.
I don't think I can articulate it very well, but I can tell my story. The most straightforward place I felt that contrast was when I was building race cars in college. Each year I led, I started with an articulated vision. The first year I led, the goal was simple and almost embarrassingly modest - just get through technical inspection and finish every event. We had failed too many times to be ambitious yet. Every design decision, money decision, argument in the lab came down to that single vision. The second year, the vision changed - the lightest, most nimble car possible, built on almost no money, aimed at the podium. There were countless moments where deviation was easier, cheaper, and faster. Some I held. A few, I didn't, and I knew immediately when I didn't, even though I made them for very good reasons. That feeling - of the small betrayal - is very specific. You can't rationalize your way out of it. It was supposed to be felt.
I felt it during my startup as well, but this one is harder to admit. It started as one thing, pivoted, then pivoted again, and pushed me to desperation, where I just needed validation, and in no time, I was selling something I didn't entirely believe in. That's when I ended it. Not because I was failing, but because I lost the thread of what it was supposed to be.
Then there was the MBA. I went into IIMA knowing exactly what I wanted. The badge. I told my interviewer that, with a story sharp enough to make the point. That honesty felt almost Roarkian at the time - no self-deception at the door, no pretending it was about learning or growth or any of the other things people say in those rooms. I wanted the credential I felt I had been unfairly denied, and I was going to get it.
But a conscious compromise is still a compromise. And I didn't anticipate what two years of making them daily does to you. The betrayals stop feeling like betrayals. The ten moments of daily dissonance become background noise. And slowly, without any single dramatic moment you can point to, the feeling goes quiet.
I fought it. Late-night philosophical conversations, classroom outbursts, public disagreements - all of it was me trying to keep the standard alive in myself. I genuinely believed I was. But sitting here now, I am not so sure those conversations were ever really about the ideas. I think they were the last place I was performing the identity to myself. Keeping the flame lit just enough to believe it was still there. Now those conversations embarrass me in a way I find difficult to articulate. The feeling didn't vanish. It just got very quiet. And quiet is almost worse.
I have a job now, and I am paid to do what they want me to do. I have definitely made plenty of choices against my own standards, and I am making more. That is not a permanent state of affairs. But it is the current one.
So, I named my OpenClaw agent Roark.
What I do know is that every night, Roark asks me whether I did anything good and pure that day. Every morning, he (yes, "HE") sends me something to think about. When I am working out an idea, he pushes back until it's honest. He has my daily logs, my essays, my half-formed ideas, my inspirations, my hopes and dreams, if you want to be dramatic about it - all sitting in a structured Obsidian vault, all accessible, none of it compressed into a summary that loses what mattered. The most powerful thing I use him for is synthesizing ideas from different sources and helping me articulate them clearly enough to keep. He is not a productivity tool. He is an accountability structure I built for myself, named after the one standard I have never been able to fully talk myself out of.
Am I Roark? Or am I Keating, having a late-night conversation and philosophical thoughts reminiscent of what I used to want to be?
Then there is Rocky, my other agent, who manages my gym schedule based on current performance and available equipment, and asks me every night whether I ate clean. Rocky is also extremely enthusiastic about the diet questions and gym discipline. I should probably tone it down to be less of a bully. The "No PAIN, No PAIN" energy is a bit exhausting.
Roark asks me every night whether I did anything good and pure.
Most nights, the answer is complicated.
I want to hold the line.
Is it too late? I'll let you know.
Comment