The 5arz Protocol.

Abstract.

Two trillion dollars of US consumer debt sits across 200 million Americans. Tens of billions of dollars are paid annually by AI labs for human-trained data, with most of the value captured by intermediaries. 5arz is a closed-loop economic system that retires consumer debt by routing the AI data economy directly to the people whose attention created it.

The mechanism is a three-step closed loop: 5arz pays a member's qualifying debt up front to the original creditor; the member completes validated AI training tasks on the 5arz job board; each task earns Stars — a 1:1 USD-pegged utility credit that settles only against the member's 5arz balance. When the balance reaches zero, the member retains all subsequently earned Stars as cash payouts. No interest accrues, ever; balances cannot increase; members may pause without penalty.

This document specifies the protocol, the multi-agent operating system that runs it (the 5arz Engine), the unit economics, the compliance posture, and the roadmap.

The double bottleneck.

Two markets are broken at the same time, in opposite directions.

2.1Consumer debt is liquidity-starved.

The average US household with revolving credit carries $7,951 in credit card balances at a weighted APR above 22%. Minimum-payment math implies an 11-year payoff on an $8,400 balance, with $14,750 in total interest paid. Existing remedies — settlement, consolidation, bankruptcy — extract value, damage credit, or take years.

The deepest problem is structural: consumers in debt are revenue-positive for lenders, so the system has no incentive to graduate them. The only escape mechanism is income — and income is the very thing under stress.

2.2AI labs are judgment-starved.

Frontier AI labs spent more than $90 billion on human-trained data in the trailing twelve months. The bottleneck for every leading model is not compute or capital — it is the supply of high-quality human judgment for RLHF, evaluation, red-teaming, multilingual coverage, and edge-case curation.

Most of this spend is intermediated. Consulting firms, data-labeling vendors, and crowdsourcing platforms capture the margin. The end worker — usually paid hourly through a vendor — sees only a fraction of what the lab pays per task, and never sees the lab.

The two markets are made for each other. The supply side of the AI data economy is the demand side of debt relief. Nobody has connected them.

The 5arz solution.

5arz is the closed-loop bridge between these two markets. Mechanically:

1. Onboarding. A member completes a soft credit pull (no score impact) via Method Financial. 5arz quotes a coverage offer for qualifying debts. On acceptance, 5arz pays the original creditor in full or in negotiated part within one business day.
2. Obligation transfer. The original credit obligation is closed. The member now holds an interest-free obligation to 5arz, equal to the amount paid. Balance never grows. Pausing is free.
3. Earning. The member completes validated AI training tasks on the 5arz job board. Tasks are sourced from contracts with frontier AI labs and labeled-data buyers. Each task is paid in Stars — a 1:1 USD-pegged utility credit.
4. Settlement. Every Star earned is automatically settled against the member's balance. When balance = 0, additional Stars convert to USD cash via Stripe payouts.

The Stars protocol.

Stars are the unit of value in the 5arz economy. The protocol is designed to be useful for members and boring for regulators.

4.1Properties.

4.2Why these properties matter.

Most "tokenized" rewards programs make a critical mistake: they let the unit float, trade, or aggregate value beyond its utility. That is the moment a token becomes a security under the Howey test, and the moment a payment program becomes a money transmitter.

Stars are designed to fail every prong of Howey:

• No investment of money — earned through labor, not purchased.
• No common enterprise — settles bilaterally between member and 5arz.
• No expectation of profit — pegged 1:1, never appreciates.
• No reliance on others' efforts — value comes entirely from the member's task work.

This is the deliberate trade-off. We give up the optionality of a tradeable token to gain regulatory clarity, member trust, and a system that won't get shut down.

The 5arz Engine — agent architecture.

5arz operates a multi-agent system to handle marketing, distribution, payment routing, and operational orchestration. The Engine is dogfooded internally and licensed externally as Agentic-as-a-Service.

5.1Topology.

Nine specialized agents organized in three functional layers, coordinated by a Conductor process and gated by Human-in-the-Loop checkpoints.

┌──────────────────────────────────────────────────┐
│              CONDUCTOR (LangGraph)               │
└────────┬─────────────┬─────────────┬─────────────┘
         │             │             │
   ┌─────▼─────┐ ┌─────▼─────┐ ┌─────▼─────┐
   │   BRAIN   │ │   MAKER   │ │   HYPE    │
   ├───────────┤ ├───────────┤ ├───────────┤
   │  Scout    │ │  Sparks   │ │ Citation  │
   │ Watchtower│ │  Render   │ │  Magnet   │
   │  Compass  │ │   Echo    │ │ Treasury  │
   └─────┬─────┘ └─────┬─────┘ └─────┬─────┘
         │             │             │
   ┌─────▼─────────────▼─────────────▼─────┐
   │       CONTEXT BUS (Pinecone + PG)     │
   └───────────────────────────────────────┘
                       │
              ┌────────▼────────┐
              │   HITL GATES    │
              │ Brand · Spend · │
              │   Sensitive     │
              └─────────────────┘

5.2Communication protocol.

Every agent exposes a standard interface: {intent, inputs, outputs, dependencies, status}. The Conductor reads agent state from the Context Bus and dispatches the next runnable. Outputs are written back to the bus with provenance metadata.

The Brand Architect agent (Compass) holds veto authority over any artifact intended for public distribution. The Treasury agent (autonomous wallet on Skyfire / Nevermined / x402 rails) holds spend authority within hard daily caps and a category whitelist.

5.3Why nine agents and not one.

A single large model can do most of these tasks. We chose specialization for three reasons: (1) separable evaluation — each agent has a measurable KPI; (2) failure isolation — a misbehaving Sparks does not corrupt Treasury; (3) independent improvement — a better hook generator can be swapped without touching the rest of the system.

Unit economics.

Per validated task completed, the cash flow looks approximately as follows:

This produces a defensible 15% gross margin while delivering 65% of the lab's spend directly to the member's debt — substantially higher than the ~30–45% pass-through typical of incumbent labeling vendors.

For a $8,400 average member balance and a 65% Stars-credit ratio, the cash needed from labs to retire the balance is approximately $12,923. At a typical task value of $3.50 paid to the member, this implies ~3,692 tasks. At 5–8 hours per week and 12-minute average task duration, this projects to 7–11 months of part-time work, which matches our beta cohort empirical data.

Compliance & regulatory framework.

The category 5arz operates in is heavily regulated. We treat compliance as the primary moat, not an afterthought.

7.1Regulatory exposure.

• FTC Telemarketing Sales Rule (TSR), Debt Relief Provisions. Disclosure and advance-fee restrictions apply to debt-relief offers.
• State Debt Adjuster Licensing. Approximately 25 states require registration or licensure. 5arz pursues licensure on a rolling state-by-state basis; service is geographically gated to licensed jurisdictions.
• CFPB authority. Applies to debt-collection-adjacent activity. 5arz is not a debt collector but follows CFPB's UDAAP standards for member-facing communications.
• Securities classification (SEC). Stars are structured as utility credits, not securities. Section 4.2 details the Howey analysis. Counsel attestation on file.
• Money transmission. Closed-loop settlement against debt held by 5arz does not constitute money transmission under most state frameworks. Bank partner facilitates creditor payouts; cash payouts are processed via Stripe.

7.2Member-facing posture.

All member-facing communications follow a strict honesty framework. We do not promise guaranteed timelines, do not imply credit-improvement outcomes, do not market the platform as "free money," and disclose the work requirement prominently.

Roadmap.

Q2 ’26Closed beta.

• 2,800-member closed beta, geographically gated to TX, FL, NY (initial state coverage).
• Three lab partners contracted at task volumes between 50K and 500K validated tasks per month.
• Stars infrastructure live; daily proof-of-reserve operational.

Q3 ’26Public launch + 5arz Engine GA.

• Expand state coverage to 12 states, target $25M debt under management.
• 5arz Engine general availability for B2B customers.
• Public API for the Engine; first three white-label deployments.

Q4 ’26Data co-op.

• Opt-in member data co-op — anonymized, member-permissioned datasets sold to AI labs as a labeled-data product.
• Members earn additional Stars for opted-in contributions.
• Daily proof-of-reserve audit; first independent quarterly attestation published.

2027Federal posture, international expansion.

• Pursue federal-level harmonization where possible to reduce per-state operational drag.
• Pilot UK and Canadian member operations through local partner banks.

Risks & open questions.

We will not pretend this works without trade-offs. The following are the live risks we monitor.

• Lab spend volatility. AI lab data spending is large but not contractually guaranteed. We mitigate via multi-lab diversification and per-task pricing escalators.
• State licensing drag. Per-state licensure is the rate-limiting step on geographic expansion. We allocate dedicated regulatory operations staff.
• Token reclassification risk. A change in SEC guidance on utility tokens could reshape Stars structurally. We hold counsel on retainer with quarterly re-attestation.
• Member churn. Members who stop completing tasks pause their balance. Long pauses degrade unit economics. We mitigate via low-friction tasks, mobile-first UX, and weekly Star earnings ceremonies.
• Adverse selection. Members with the highest debt-to-judgment ratio self-select, which compresses lab payouts. We mitigate via task quality scoring and tier-based access to higher-value work.