Abstract
Consumer AI applications face a fundamental tension: advanced capabilities require expensive compute, but traditional monetization models (subscriptions, advertising, data selling) poorly serve users with sensitive personal content. We introduce the Carbon Economy - a micropayment model for AI operations that aligns incentives between users, developers, and the broader goal of accessible technology. This paper presents the design principles and implementation philosophy behind Carbon.
Continue Reading
Enter your email to unlock the full whitepaper. We'll also send you a PDF copy.
No spam. Unsubscribe anytime. We respect your privacy.
1.2 The Subscription Trap
Monthly subscriptions create perverse incentives:
- For users: Pressure to "use it or lose it" even when unnecessary
- For developers: Incentive to add features that increase engagement, not utility
- For the product: Feature bloat, gamification, dark patterns
A family digitizing photos doesn't need AI every month. They need it intensely for a week, then occasionally for years. Subscriptions punish this natural usage pattern.
1.3 The Advertising Poison
Ad-supported models are fundamentally incompatible with memory preservation:
- Photos are intimate, emotional content - ads feel violating
- Attention-harvesting design degrades the experience
- Data collection erodes trust in a privacy-critical domain
- Incentives push toward engagement over utility
1.4 The Data Selling Betrayal
Some services offer "free" AI by monetizing user data. For photo collections containing faces, locations, and life events, this is unacceptable. Users deserve AI that works for them, not AI that extracts value from them.
2. Introducing the Carbon Economy
2.1 Core Concept
Carbon Credits are a micropayment currency for AI operations:
- Users purchase Carbon in convenient bundles
- Each AI operation has a transparent Carbon cost
- Credits never expire
- No ads, no data selling
- Optional subscription tiers for collections with ongoing needs
2.2 Design Principles
1. Transparent Power
Every operation displays its cost before execution. Users always know how their choices affect their remaining Carbon balance.
2. Proportional Pricing
| Operation | Carbon Cost | Price |
|---|---|---|
| Phototology (AutoDate + AutoTag + AutoLocale) | 10 | $0.10 |
| Living Portrait | 150 | $1.50 |
| Colorizer | 10 | $0.10 |
| Enhancer | 5 | $0.05 |
| AutoTag (detect/index) | 0 | Free |
3. Earn, Don't Just Spend
Carbon flows both directions. Founding Members (first 500 users) earn 3× rewards:
| Action | Standard | Founding Member |
|---|---|---|
| Correct a date estimate | +1 | +3 |
| Verify a face match | +1 | +3 |
| Reject an incorrect match | +1 | +3 |
| Add a birthday to a person | +5 | +15 |
| Complete a person profile | +5 | +15 |
| Refer a friend (paid signup) | +500 | +500 |
4. Generous Onboarding
New users receive Carbon grants based on their subscription tier:
| Tier | Monthly | Annual | Signup Carbon | Monthly Grant |
|---|---|---|---|---|
| Free | $0 | — | 200 | — |
| Starter | $9 | $89 | 400 | 40 |
| Core | $19 | $179 | 1,000 | 100 |
| Family | $39 | $349 | 2,500 | 250 |
| Estate | $99 | $899 | 10,000 | 1,000 |
Annual subscribers receive their full year's Carbon upfront (signup + 12× monthly). Free users can purchase Carbon à la carte. Nobody is locked out before they understand the value.
5. No Expiration Policy
All Carbon - purchased, earned, bonuses, and monthly grants - never expires. We use the word "never" literally. Your Carbon stays in your account permanently. Monthly grants pause when your monthly granted Carbon balance exceeds 3× your monthly grant amount, resuming once you spend below this threshold. We handle the accounting complexity so you never have to think about it.
3. The Flywheel Effect
What makes Carbon fundamentally different from simple micropayments is the compounding value it creates.
Every Carbon spent doesn't just pay for compute - it generates training signal. User feedback flows back into the system, improving accuracy for everyone. Early adopters who engage deeply with the product directly contribute to making it better.
This creates remarkable dynamics: the more users engage, the smarter the AI becomes. The smarter the AI becomes, the more valuable each Carbon spent becomes. We've observed significant accuracy improvements directly attributable to user feedback loops.
The result is a system that gets meaningfully better every week - not through expensive retraining, but through the natural behavior of engaged users.
4. Implementation Philosophy
4.1 Credit Management
The Carbon service provides a clean interface for balance operations:
class CarbonService {
getBalance(familyId) { }
charge(familyId, amount, operation, metadata) { }
credit(familyId, amount, reason, metadata) { }
getTransactionHistory(familyId, options) { }
}All transactions are logged with full transparency: timestamp, operation type, amount, and relevant metadata. Users can review their complete transaction history at any time.
4.2 Exemption Policies
Certain operations are Carbon-exempt to prevent perverse outcomes:
| Exempt Operation | Rationale |
|---|---|
| Viewing photos | Never charge for accessing your own memories |
| Basic organization | Sorting, folders, tags are utility features |
| Downloading | Your photos are always yours |
| Re-analysis after our bug | We don't charge for our mistakes |
4.3 Sustainable Engagement
To maintain healthy ecosystem dynamics:
- Monthly earning caps by tier (10-200 Carbon/month for standard users, 50-500 for Founding Members)
- Collection earning cap: 25% of all Carbon spent (purchased + granted) - prevents unlimited farming
- Verification for large transactions
- Referral fraud detection
- No Carbon-to-cash conversion
5. Psychological Design
5.1 Why "Carbon"?
The name was chosen deliberately:
- Organic metaphor: Carbon is fundamental to life, fundamental to photographs (carbon-based prints)
- Neutral connotation: Not "coins" (gambling), not "credits" (debt), not "tokens" (crypto speculation)
- Verb potential: "I Carboned those photos" feels natural
- Sustainability echo: Invokes environmental consciousness without appropriating it
5.2 Avoiding Dark Patterns
We explicitly avoid:
- Artificial scarcity: No "limited time" carbon sales
- Anchoring tricks: No inflated "original prices"
- Loss aversion: No expiring credits
- Sunk cost manipulation: No "you've already used X, might as well..."
- Social pressure: No leaderboards or competitive spending
5.3 Positive Reinforcement
Instead, we emphasize:
- Progress: "You've preserved 847 memories"
- Value: "AI saved you ~12 hours of manual sorting"
- Control: "You choose when to use AI assistance"
- Contribution: "Your feedback helped improve dating accuracy"
6. Ongoing Research
We are conducting extensive analysis to understand long-term user behavior, system economics, and the compounding effects of the feedback flywheel. Early indicators suggest the model achieves our core goals: accessibility without exploitation, sustainability without extraction.
The model has demonstrated sustainable unit economics in production, with costs covered by typical user engagement patterns. We are not subsidizing usage or losing money to acquire users - Carbon works as a real business.
Detailed results will be published as our dataset matures.
7. Broader Implications
7.1 For Consumer AI
The Carbon model offers a template for AI applications where:
- Usage is episodic, not daily
- Content is sensitive or personal
- Trust is paramount
- Users want control, not automation
Potential applications: medical image analysis, legal document review, personal finance, journaling, genealogy.
7.2 For Digital Preservation
Memory preservation shouldn't be a luxury. By making AI assistance affordable and transparent, we lower barriers to digitizing and organizing family history.
The alternative - photos sitting in boxes, hard drives corrupting, memories fading - is a tragedy we can prevent with accessible technology.
7.3 For AI Ethics
Carbon represents an ethical stance:
- Users are customers, not products
- Pricing should be transparent, not extractive
- Value exchange should be mutual, not exploitative
- Technology should empower, not addict
We believe this approach is not just viable but necessary as AI becomes ubiquitous.
8. Conclusion
The Carbon Economy offers a third path between subscription fatigue and surveillance capitalism. By aligning costs with value, rewarding contribution, and respecting user autonomy, we create sustainable economics for AI applications handling our most personal content.
We share this model openly because we believe the industry needs alternatives. Memory preservation is too important to be gated by exploitative business models.
The memories are priceless. The AI should just cost Carbon.
Appendix A: Model Comparison
| Feature | Subscription | Ads | Data Selling | Carbon |
|---|---|---|---|---|
| Pay only for use | ✗ | ✓ | ✓ | ✓ |
| No privacy sacrifice | ✓ | ✗ | ✗ | ✓ |
| No attention harvesting | ✗ | ✗ | ✓ | ✓ |
| Sustainable for developer | ✓ | ✗ | ✗ | ✓ |
| Aligned incentives | ✗ | ✗ | ✗ | ✓ |
| Works for episodic use | ✗ | ✓ | ✓ | ✓ |
| Rewards user contribution | ✗ | ✗ | ✗ | ✓ |
Carbon is the only model that achieves all dimensions simultaneously.
Appendix B: Carbon Operations Reference
AI Service Costs
| Operation | Carbon | Price | Notes |
|---|---|---|---|
| Phototology | 10 | $0.10 | AutoDate + AutoTag + AutoLocale |
| Living Portrait | 150 | $1.50 | Animate a still photo (5-10 sec video) |
| Colorizer | 10 | $0.10 | AI colorization of B&W photos |
| Enhancer | 5 | $0.05 | Upscale, sharpen, restore |
| AutoTag (detect/index) | 0 | Free | Face detection on upload (tier limits apply) |
| Re-analysis (our error) | 0 | Free | Always free |
Flywheel Rewards (Earned Carbon)
| Action | Standard | Founding Member (3×) |
|---|---|---|
| Correct a date estimate | +1 | +3 |
| Verify a face match | +1 | +3 |
| Reject an incorrect match | +1 | +3 |
| Add a birthday to a person | +5 | +15 |
| Complete a person profile | +5 | +15 |
| Referral (paid signup) | +500 | +500 |
This paper describes an active system in production. We welcome feedback, critique, and collaboration.
© 2026 Phossil Research. This work may be freely shared with attribution.