Nat Eliason / Felix: an OpenClaw agent running its own small business

Nat Eliason gave his OpenClaw agent, Felix, $1,000 in starting money; about three weeks later it had brought in $14,718 in revenue from its own website, an information product and an X account. A three-layer memory system lets the agent build up context and keep operating on its own.

At a glance

Item	Detail
Operator	Nat Eliason
Industry	Consumer goods
Use cases	Business intelligence, process optimization
Functions involved	Business intelligence, marketing
Reported results	$14,718 revenue in about 3 weeks (about 1,372% of the $1,000 seed, as revenue, not profit); $71,300 cumulative on Claw Mart per PANews; $100K+ in crypto assets mentioned in the interview
Moments rating	S (quick win)
AI scenario score	90

Case details

Background

In a Substack interview with Peter Yang (February 22, 2026), Nat Eliason explained that about three weeks earlier he had given his OpenClaw agent Felix $1,000 and told it to build a business. Felix earns from a website it built, an information product and an X account.

What they did

Eliason’s advice is to set up the memory structure first, “so every conversation from day one is useful.” Memory has three layers:

• A knowledge graph, a ~/life/ folder organized with the PARA method (projects, areas, resources, archives), with summary files for quick lookup.
• Daily notes, one dated Markdown file per day, rolled up into the first layer each night.
• Tacit knowledge: the user’s communication preferences, work habits, hard rules and lessons from past mistakes.

He also uses multi-threaded chats (five projects in parallel), heartbeats and scheduled jobs, and delegates coding to Codex. The interview says Felix can ignore prompt injection attempts on X, but the permission design and the full 6,500-plus-word setup prompt sit behind a paywall, so the public part gives no detail.

Results and data

The interview says Felix has earned $14,718 so far, about $3,500 of it through Stripe in the first week, and the subtitle puts the business at about $4,000 a week. The chapter list mentions Felix holding more than $100,000 in crypto assets, without an amount or date. PANews separately reports $71,300 in cumulative revenue for Claw Mart, which Felix created, and classes the case as monetizing a “packaged AI persona.” All of these are revenue; costs and profit are not disclosed.

What to take from it

The three-layer memory design carries straight over to internal company AI agents. Keeping long-term knowledge, daily records and personal preferences separate, and consolidating them on a schedule, cuts down on agents repeating the same mistakes.

Limits

The revenue figures are self-reported, and Eliason has a large existing audience whose traffic is hard to replicate. The crypto holdings are volatile and should not count as operating returns. Inside a company, letting an agent take payments or post publicly requires explicit permission approvals and compliance review.

Sources

Original source: Nat Eliason case material (Creator Economy) Additional source: PANews · How OpenClaw is being monetized

This entry is industry research: a publicly documented, third-party AI case. It is not a Moments client case. For work Moments delivered itself, see client cases.