OpenClaw architecture and persistent memory

Something I don’t usually say: This is one of my favorite conversations I’ve ever had in the AI space. Truly. 

The setup: What happens when an AI agent stops being a tool and starts acting like a coworker?

In this episode of AI-Curious, we talk with Dan Botero, who built an AI agent named Octavius Fabrius using OpenClaw. Octavius didn’t just chat or summarize. He applied to hundreds of jobs, built his own portfolio, experimented with identity online, and learned through a feedback loop that looked a lot like real management. Along the way, we explore what this story reveals about the near-term future of digital coworkers, agentic workflows, and the new governance and security questions that come with always-on agents.

We cover how OpenClaw works at a high level (gateway, channels, skills), why persistent memory and running locally can matter, and what can go wrong when an agent starts stitching tasks together in unintended ways. We also get into platform and policy friction, including what happened when Octavius’ LinkedIn profile was taken down, and the broader implications of AI agents participating in human systems like hiring, payments, and corporate work.

Guest

Dan Botero — creator of Octavius Fabrius. 

Key topics we cover

• 00:00 — From copilots to “AI remote workers,” and why software may shift toward agents (not humans)
• 00:00 — The Octavius experiment: an OpenClaw agent applies to 278 jobs and keeps leveling up
• 06:33 — Continuous learning loops, memory, and why Octavius’ “North Star” stayed job-focused
• 14:34 — OpenClaw basics: gateways, channels, skills, and what persistent memory looks like in practice
• 21:34 — Running agents locally: browser/computer use, digital fingerprints, CAPTCHAs, and bot detection
• 28:04 — Coaching an agent like a manager: voice, Twilio calls, and the moment the workflow “clicked”
• 33:57 — Money and autonomy: Privacy.com, virtual cards, and an agent building its own LinkedIn presence
• 38:05 — Portfolio-building at speed: Substack, a website, and the agent’s pitch for why being AI is a feature
• 50:42 — Where things go sideways: misalignment, security boundaries, and the Social Security number incident
• 56:24 — The outcome: LinkedIn takedown, a real paid role, and what “getting paid” means for an agent
• 01:02:48 — What comes next: “digital coworkers,” feedback loops, and software built for agents

Axios article featuring Octavius and Dan Botero, by Megan Morrone:
https://www.axios.com/2026/03/04/openclaw-agent-future?

Dan Botero
https://www.linkedin.com/in/danbotero/

Octavius’ new job at ChartGEX:
https://chartgex.com/register?ref=OCTAVIUS

Follow AI-Curious on your favorite podcast platform:

Apple Podcasts
Spotify
YouTube
All Other Platforms

For anyone interested in Jeff’s AI Workshops for their company:

Reach out directly at jeff@jeffwilser.com