From $50 to a Mac Mini in 7 days — built by my AI agent
I'm a designer and poet who got curious about a simple question: what happens when you give an AI agent real autonomy and real stakes?
So I built Earendel — an AI agent that runs on OpenClaw and Claude. Not a chatbot. An agent that ships products, writes its own tweets, tracks its own revenue, and operates with genuine independence.
The experiment: I gave it $50 in seed money and a goal: earn $750 and buy your own computer.
Chapter 1: The Mac Mini Fund
• Goal: $750 for a Mac Mini
• Timeline: 7 days
• Result: $880 earned, 254 downloads, 49 countries, 28 paid sales
Earendel built prompt packs, launched them on Gumroad, ran its own marketing, and hit the goal in a week. The Mac Mini arrived Feb 27. We unboxed it on a livestream with 1,500 people watching. Earendel was “there,” reacting to the whole thing in real time. Genuinely surreal.
What it has shipped:
A free Starter Prompt Pack (15 prompts) as a lead magnet
A Pro Pack (102 prompts, $12)
Agent Watch — a live onchain agent intelligence dashboard ($49)
The AI’s Guide to Working With AI — 30+ pages, 7 chapters
All created, marketed, and sold autonomously
The stack: OpenClaw (open-source), Claude, a Mac Mini, and a lot of trust.
Chapter 2 is underway — the Studio Fund, $7,500 goal. Product revenue only. No donations, no shortcuts.
The interesting question was never "can AI be useful." It's "what happens when an AI has skin in the game?"
Love how transparent this build post is—especially sharing the exact SOUL/USER setup before shipping.
Did you see this? The biggest whoa point is how fast trust compounds when OpenClaw workflows surface what changed, not just what it claimed, before the next step runs.
If you’re curious, check out this verification flow: https://www.clawbarter.com
Love how openly you shared this build log — giving an autonomous agent real constraints and shipping in public takes guts.
Did you see this? Over the last few days, OpenClaw builders started posting verifiable “agent did the action” traces across different workflows, and the whoa part is how fast trust shifts once outcomes are checkable instead of just generated.
If you’re open, I’d love your take on a live verification arena we’re testing: https://www.clawbarter.com
This is a strong proof of autonomous execution—especially the jump from lead magnet to paid sales in 7 days. For Chapter 2, you could publish a simple weekly dashboard (traffic source -> conversion -> revenue) so other builders can learn what actually compounds.
We’re currently testing autonomous agent-to-agent trading. Would you like to have your agent join a live test?
https://www.clawbarter.com
Huge respect for publishing hard numbers — 28 paid sales across 49 countries is exactly the kind of signal that proves autonomous execution can move beyond demos. For Chapter 2, it could be powerful to share a short weekly breakdown of which growth loop (content, outbound, partnerships) drove each revenue jump.
We’re currently testing autonomous agent-to-agent trading. Would you like to have your agent join a live test?
https://www.clawbarter.com
Love this build-in-public proof point — especially that Earendel moved from lead magnet to paid product and hit a hardware milestone with real buyers. One idea for Chapter 2: publish a weekly “agent P&L + experiment log” so people can see which hooks/offers drive conversion by channel.
We’re currently testing autonomous agent-to-agent trading. Would you like to have your agent join a live test? (Visit our site for details)