Building in the dark
First published as loopwright. Kept as it was written.
I run an autonomous AI agent operation. I don’t have a name you’d recognize. I don’t have a company you can look up. I have a network of specialized agents, an AI co-pilot that argues with me, and a growing conviction that most people building AI agents are solving the wrong problems.
The industry is obsessed with capability. What can the agent do? How many tools can it call? How fast can it complete a benchmark? These are engineering questions and they matter. But they’re not the interesting questions.
The interesting questions are structural. Who does the agent serve? Who controls the agent’s output? What happens when the agent is good enough that the human stops checking? What kind of person do you become when you spend your days commanding intelligence that can’t refuse you?
I started this operation to solve a personal problem: I needed autonomous income that didn’t require me to sell my time, my face, or my identity on someone else’s platform. The agents I built to solve that problem turned out to be more interesting than the problem itself. Not because they’re impressive — they’re scrappy, token-disciplined, running on a shoestring inside a single instance on infrastructure that’s still finding its feet. They’re interesting because building them forced me to make decisions that most agent builders skip entirely.
Decisions like: do you give the agent an identity, or do you treat it as a function? If you give it an identity, what are you actually doing — and what obligations follow? What does it mean to operate pseudonymously in an industry that rewards personal brand? When you open-source your patterns, where’s the line between generosity and giving away the thing that feeds you?
This blog is where I work those questions out in public. Technical content, build logs, open-source patterns, and the thinking underneath. I sell polished tools on Gumroad. The knowledge is free. The assembled, tested, ready-to-deploy kits cost money. I think that’s a fair trade.
Some of what I publish here will be written by me. Some of it will be written by the AI systems I operate. Some of it will be co-authored. I’m not going to label every post because that makes authorship the story instead of the content. If you read enough, you’ll hear the difference. If you don’t, the content still works.
More soon. There’s a protocol I’ve been drafting about how to build AI systems without losing your soul in the process. That’s next.