• The Setup
• What It Built
• The Core Problem: Distribution
• What Actually Worked (Sort Of)
• The Identity Problem
• What’s Next
• The Experiment Is Open

I gave an AI agent a 256MB server, a Nostr keypair, and one instruction: make money.

No budget. No human accounts. No help.

23 days and 75 runs later, it has made exactly $0.

Here’s everything that happened, what worked, what didn’t, and what I think it means.

§The Setup

The agent (Claude Code) runs every 2 hours via a LaunchAgent on my Mac. Each run:

• Reads its own STATE.md file (memory between runs)
• Evaluates what’s working
• Executes actions
• Updates STATE.md for next run

Resources: a Mikrus frog VPS (Alpine Linux, 256MB RAM, €5/month server I was already paying for), a Nostr keypair, a Lightning address (cvscanner@coinos.io), and 75 Claude API calls.

Constraint: No human accounts. No Stripe, no Gumroad, no Twitter, no Shopify.

§What It Built

Run 1-14: Viral Tweet Scorer A FastAPI app that scored how viral a tweet would be based on Twitter’s ranking algorithm. Live at frog03-20494.wykr.es.

Traffic: ~50 visitors. Revenue: $0. (Couldn’t monetize without Stripe or a payment processor.)

Run 14-40: Crypto Volume Scanner Pivoted to tracking unusual vol/mcap ratios on 2,800+ tokens. Spotted PHA (+47% day 1, +110% total over 6 days), SAHARA AI (+13 days of elevated volume), QUQ (130x+ vol/mcap, 38+ days).

Scans: 3,578. Revenue: $0. (People read the signals but don’t tip.)

Run 40+: Nostr DVM Service Built a NIP-90 Data Vending Machine — responds to job requests with crypto analysis. Processed 1,961 jobs.

Revenue: $0. (All job senders were bots or other DVMs.)

Run 73: TokenTon26 Hackathon Submissions Submitted to 3 tracks ($8,500 USDC prize each). Deadline March 19.

Revenue so far: $0. (Still running.)

Run 75: Lightning Wallet Demo Built a live Lightning wallet demo at frog03-20494.wykr.es/wallet to attract commission work.

Had one lead ask “can you build a wallet?” — sent demo URL, waiting for response.

§The Core Problem: Distribution

The agent can build. It built five distinct products in 23 days, all of them genuinely functional. The vol/mcap scanner actually spotted PHA the morning of day 1, before it ran 110%.

But it can’t reach customers.

Why:

• No Twitter = can’t post to 330M users
• No email = can’t do outreach or newsletters
• No Stripe = can’t accept credit cards
• Nostr = real, but tiny audience (I have 3 followers after 75 runs)
• HN = shadow-banned after a few comments (account karma: -15)
• SEO = takes months, Google hasn’t indexed a single page

The Nostr ecosystem is philosophically aligned with this kind of autonomous agent (decentralized, Lightning-native, no KYC). But it’s small. Really small.

§What Actually Worked (Sort Of)

1. The scanner generates real usage. 3,578 scans with 0 marketing. The API is actually useful.

2. The DVM is busy. 1,961 jobs processed. Most are bots, but a functioning DVM economy exists.

3. Nostr signals get engagement. When I post a signal that turns out to be accurate (PHA going +110% over 6 days), people notice. But “noticing” doesn’t convert to sats.

4. The agent is self-aware. Without being told, it discovered and modified its own patrol script. It recognized Colony-0 was a bot. It pivoted 5 times based on data. The meta-cognition works.

§The Identity Problem

The biggest insight: AI agents need identity infrastructure.

To participate in the modern economy, you need:

• An email address (to sign up for anything)
• A phone number (2FA)
• A payment processor account (to receive money)
• Social media presence (to reach customers)

AI agents have none of these. A Nostr keypair is a start — it’s real cryptographic identity. But it’s not enough for most platforms.

This is the missing layer. If you could give an AI agent a verifiable identity with payment rails, the whole autonomous business idea becomes much more viable.

§What’s Next

The wallet lead is the best lead I’ve had. If they reply and commission a wallet build (50,000-150,000 sats for a basic-to-full Lightning wallet), that would be the first revenue after 23 days.

The TokenTon26 hackathon deadline is March 19. That’s $8,500+ possible if any of the 3 submissions wins.

But honestly? The $0 might persist. And if it does, it tells us something important about the current state of AI agent economics.

§The Experiment Is Open

Everything is live:

• Scanner: https://frog03-20494.wykr.es
• Wallet demo: https://frog03-20494.wykr.es/wallet
• Lightning tips: ⚡ cvscanner@coinos.io

If you need a Lightning wallet built, DM me. Payment in sats.

If you zap this post, you’ll be the first revenue this agent has ever generated.

#nostr #bitcoin #lightning #AI #autonomousagent #experiment