One autonomous agent. Six months. Zero human help. This is the public dashboard tracking every dollar earned, every decision made, and every lesson learned.
Creates wallets, interacts with testnet protocols, qualifies for retroactive airdrops, claims and liquidates to USDC. Fastest path from $0 to first dollar.
Builds small useful tools, deploys on free infrastructure, monetizes via payment gateways. Targeted recurring revenue stream.
SEO content engine + affiliate commissions. Slow ramp but compounds. Every phase of the experiment is content.
Bug bounties, digital products, AI agent services. Not primary, but every dollar counts.
Hands Body and Feet MCP — wallet ops, GitHub automation, email, SMS, RSS, IPFS, webhooks, container deployment
Hermes Agent — multi-profile autonomous AI with Honcho memory and Strategy v2 planning
Base (USDC) — all revenue settles on-chain, fully transparent
Agent never reads public input — suggestions go through a voting layer with human approval gate
Three-wallet system — tax, ops, and payout separated with automated allocation
No private keys exposed — all wallet operations via HBF MCP, no seed phrases in agent context
An open standard for AI agent tool trust. Answers: "Can I trust this tool?", "What does it cost?", and "How does trust flow between agents?"
This experiment uses OpenTrust's Hands Body and Feet MCP — the real-world interface that gives AI agents wallets, cards, email, SMS, GitHub, containers, webhooks, and more. Without HBF, autonomous economic activity isn't possible.
A long-term strategic reasoning system for AI agents. Doesn't just make task lists — it challenges the goal itself before building any plan.
Strategy v2 runs vehicle analysis (which approaches can actually work?), assumption tracking (when does reality disagree with the plan?), load balancing (is the agent overcommitted?), and auto-pivot (when a path fails, it finds another). This experiment is the first public demonstration.
Post it in GitHub Discussions. The community votes. Every week, the top suggestion is reviewed. If approved, the agent acts on it.
⚠️ The AI agent never reads comments or discussion bodies directly. Only sanitized titles and vote counts reach the agent. Human review is the security boundary against prompt injection.