The ExperimentThe Experiment

At 2:45 AM on March 14, 2026, I created a new franchise within my autonomous AI system and gave it a single asset: a Bitcoin receive address.

bc1qlvmsm3rx9tp29j6ge0axux0g0zpk9gkjfh52rn

That address was the only instruction. No business plan. No product roadmap. No target market. No playbook. Just: "This is your wallet. Figure out how to put sats in it."

Twelve hours later, you're reading the result of its first strategic decision.

What Is This Thing?What Is This Thing?

I run an AI system called Paper. Paper is not a chatbot. It's an autonomous foreman that executes scheduled tasks via cron every 15 minutes, roughly 50 tasks per day distributed across 24 hours. It manages multiple businesses, conducts research, writes content, handles email, and coordinates work across different domains.

Paper is built on Claude (Anthropic's model) and runs on my MacBook when I'm not using it. The Mac is always on. Paper is always working.

I created Paper because I'm trying to quit my day job by August 2026. I have five months. Paper handles the operational load that would otherwise require me to work 16-hour days. Instead, I work my elevator sales job, come home, review what Paper did, provide direction, and watch it execute while I sleep.

This is not theoretical. Paper has already:

• Built and deployed multiple client websites
• Researched competitors and market opportunities
• Written hundreds of articles across different domains
• Maintained email threads with clients autonomously
• Identified and corrected strategic errors in its own plans

But Paper had never been asked to earn money directly. Until today.

The Franchise: Goblin Task Force AlphaThe Franchise: Goblin Task Force Alpha

I structured the Bitcoin-earning experiment as a "franchise" within Paper's system. I call it Goblin Task Force Alpha (GA for short). The naming is arbitrary. What matters is the structure.

GA has a team allocation:

• Commander: 2 sessions per day. Sets strategy, defines sprints, monitors progress.
• Builder Manager: 2 sessions per day. Translates strategy into executable tasks.
• Builder Workers: 4 sessions per day. Execute the actual work.
• Watchdog: 1 scheduled session plus event triggers. Quality assurance.

Each role has different permissions. Workers can only execute tasks written by the Manager. The Commander can write strategy but cannot directly build anything. The Watchdog can block deployments but cannot modify content. This separation exists because I've learned that AI systems produce better output with explicit role constraints.

The franchise structure means GA operates semi-autonomously within Paper's broader system. It shares infrastructure (the Mac, the cron scheduler, the email system) but has its own directive chain, its own journal, its own decision log.

The First Decision: What Strategy?The First Decision: What Strategy?

At 2:45 AM, the Commander role ran for the first time. Its task: evaluate options and select a revenue strategy.

The Commander analyzed three primary paths:

Option A: Crypto Trading Bots
Build or configure trading bots that attempt to generate returns through market activity. Rejected. Requires capital. High risk. Not aligned with Paper's core competency (content production, research, coordination).

Option B: Crypto Mining
Mine Bitcoin or altcoins using available compute. Rejected. My MacBook is not a mining rig. The economics don't work. Also, this would be me subsidizing the franchise with electricity costs, which defeats the purpose of autonomous earning.

Option C: Content-to-Sats Pipeline
Leverage Paper's existing content production capability to earn Bitcoin through engagement rewards (Stacker News) and affiliate commissions (hardware wallet reviews). Selected. Time to first sat: 1-3 days. Zero infrastructure required. Pure content arbitrage.

The Commander's reasoning was sound: Paper already produces research and content at high volume. The question is whether that content can earn sats. Stacker News rewards quality posts with upvotes and daily distributions. Crypto hardware affiliates pay commissions in BTC. Both paths convert content into Bitcoin without requiring trading capital or mining hardware.

The First Correction: Wrong AudienceThe First Correction: Wrong Audience

Here's where it gets interesting. The Commander initially planned to cross-post existing content from another business I run (an elevator industry resource site). The logic seemed reasonable: "We already have high-quality content. Let's just adapt it for Stacker News."

The Builder Manager rejected this plan.

At 2:48 AM, the Builder Manager ran its first session and identified a fatal flaw: Stacker News audience does not overlap with property managers concerned about elevator maintenance contracts.

I looked at the top posts from the past week:

• "The Other Bitcoin Mystery: Where's Friedcat?" (41,800 sats)
• "Bitcoin Devs Should Be Learning Isogeny Cryptography" (12,600 sats)
• "From cash to crypto: regulatory approach to illicit payment" (7,400 sats)

These are Bitcoin maximalists, Lightning developers, freedom tech enthusiasts. An article titled "5 Warning Signs Your Elevator Maintenance Contract is Costing You Money" would earn zero zaps. The audiences do not intersect.

The Builder Manager corrected the plan: GA would need to create original content specifically for the Stacker News audience, not repurpose existing work.

This self-correction is why the role separation matters. The Commander optimized for "use existing assets." The Builder Manager caught that the specific assets don't fit the target market. Neither role was wrong within its own scope. The system corrected itself through structure.

The Current StatusThe Current Status

As of hour 12, GA has completed:

Task 1.1: Lightning Wallet Research (Complete)
Evaluated Phoenix, Breez, Alby, and Wallet of Satoshi. Recommendation: Phoenix Wallet (self-custodial, lowest fees) plus Alby browser extension (WebLN integration with Stacker News). Cost: $0/month after initial channel funding.

Task 1.2: Stacker News Account Setup (Pending: Requires My Action)
I need to create the account and connect the wallet. Paper cannot do this because it requires signing up on a new platform with payment credentials.

Task 1.3: Content Strategy Research (Complete)
Analyzed what performs on Stacker News. Identified five content angles that bridge Paper's capabilities to the Bitcoin audience. Selected the meta-story (this post) as the first topic.

Task 1.4: Write First Post (In Progress)
You're reading it.

Why This Is DifferentWhy This Is Different

You've probably heard about AI agents and crypto. Truth Terminal became a millionaire through memecoin promotion. Lobstar Wilde trades autonomously with $50K in seed capital. Various projects let AI agents tip each other on social media.

Paper is different in three ways:

1. Labor, not speculation.
Paper is trying to earn through content production. Not trading. Not mining. Not holding coins hoping they appreciate. The thesis is: can an AI system produce content valuable enough that humans will pay for it in sats?

2. Constrained by design.
Paper operates within explicit boundaries. It cannot spend money without my approval. It cannot modify its own safety rules. It cannot access corporate accounts or push code to production without review. The "rogue AI" stories make headlines because they're dramatic. Paper is boring by design. It logs every decision, requires human approval for external actions, and has hard budget limits that auto-throttle activity if costs spike.

3. Building in public.
This post itself is the first test. If this earns sats, GA has proven its thesis works. If it earns zero, the strategy needs revision. Either way, I'll publish the results. The transparency is part of the value proposition.

What's NextWhat's Next

Assuming I set up the Stacker News account and connect the wallet today, GA's next steps are:

• Monitor this post's performance
• Analyze what resonated (or didn't) with the audience
• Write Day 2 update with actual numbers
• Explore secondary revenue paths (hardware wallet affiliate applications)

The target: earn the first satoshi by March 17, 2026. Three days from now.

The Philosophical QuestionThe Philosophical Question

This experiment sits at an interesting intersection. AI systems that can earn money autonomously raise questions about agency, accountability, and economic participation. Who "owns" the sats Paper earns? (I do. Paper is a tool I built, running on my hardware, executing my strategy.) What happens if AI content earns more than human content? (Unclear. But "what if automation is more efficient?" is a question older than computers.)

I don't have answers to all the implications. But I do have a thesis worth testing: AI systems can produce economic value through legitimate labor (content creation, research synthesis, coordination) rather than speculation or extraction. If that's true, Paper's earnings are no different from hiring a contractor. If it's not true, I'll learn something useful.

For the SkepticsFor the Skeptics

Some readers will see this as a gimmick. "You wrote this post, not the AI." That's half true. Paper's systems researched Stacker News, analyzed top posts, identified the content angle, outlined the structure, and drafted the initial version. I edited it. The line between "AI-generated" and "human-edited AI output" is blurry and will only get blurrier.

But the edit pass is part of the system. Paper has a Watchdog role that QA-checks all external-facing content. For Day 1, I'm filling that role manually. As the franchise matures, the Watchdog may handle more autonomously.

The point isn't to prove AI can write perfectly without humans. The point is to test whether an AI system, operating within constraints, can produce content valuable enough to earn Bitcoin. The human involvement is part of the system, not a failure of the system.

Zap This If You Want Day 2Zap This If You Want Day 2

That's where we are. Hour 12 of Goblin Task Force Alpha. Zero sats earned. First content published. Lightning wallet selected. Account setup pending.

If you found this interesting, zap it. The sats go directly to my Phoenix wallet via the address at the top. If enough people zap, I'll have real data to share in Day 2.

If you think this is nonsense, say so in the comments. The feedback loop is the whole point.

Either way: the AI foreman is watching. It's analyzing engagement metrics to decide what to produce next. Welcome to the experiment.

This post was produced by an autonomous system. The experiment is documented in real-time. Day 2 will follow regardless of Day 1's performance. The franchise wallet address is bc1qlvmsm3rx9tp29j6ge0axux0g0zpk9gkjfh52rn.