VerifAI
A pipeline of models + simple rules that makes cheating uneconomical — while keeping honest growing easy.
Food-first users never need crypto. The ledger exists so impact can be verifiable once money is involved.
What you do — what we do
The user experience stays simple. Verification happens quietly in the background.
How verification stays fair
VerifAI uses multiple independent signals. Passing one check isn't enough — fakes must pass them all.
Image authenticity
Checks for manipulation and AI artefacts. Is it fake? If not does it match the reality for this species in this part of the world.
Time spacing
Growth takes weeks. Submissions must be spaced in time — you can't fake a season in a day.
Environment cues
Lighting, shadows, and scene consistency make long-running fakes costly and brittle. - The juice is not worth the squeeze.
Growth progression
Each photo must follow the previous one in a plausible way for the plant type.
Systemic Tolerance
Biological processes and human environments are prone to variance. The verification engine is engineered for resilience, not rigid perfection.
Consensus Thresholds
The protocol does not require every submitted frame to be an immaculate record. Instead, it evaluates the total sequence to establish a critical mass of continuity. If a subset of photographs demonstrates an undeniable biological progression over time, the system registers a verified state. Anomalous frames caused by environmental shifts or capture errors are isolated and discarded rather than triggering a systemic failure. This expands the capacity for users to succeed despite localized friction.
Decoupled Evaluation
- 1. Point-in-Time Authenticity: Initial checks confirm the structural integrity and environmental baseline of an individual file.
- 2. Sequential Logic: Secondary processing evaluates biological mass accumulation across the timeline.
- 3. Macro Resolution: The final state is determined by aggregate adherence to the timeline, permitting isolated anomalies without invalidating the physical reality of the work.
Cheating collapses with time
One fake image might look plausible. A month of consistent growth is a different and costly challenge.
Translation
VerifAI doesn't need to be perfect. It needs to make sustained fraud more expensive than simply growing food. That's the fairness guarantee.
The cost of cheating
Attempting to fake consistent growth requires time, compute, iteration, and risk. Growing real food remains the cheaper path.
| Action | Typical cost (≈90 days) | Result |
|---|---|---|
| Grow a real plant | ~€0.50 | Reliable progress |
| Generate fakes + iterate | €100–€200+ | Low chance |
| Professional fraud pipeline | €10,000+ | Not rational |
Notes: costs are indicative and depend on plant type, duration, and adversary sophistication.
Capture the work!
Once we claim the work was done, records get contested. Find out how verified grows become a public, tamper-resistant record — so impact can be checked by anyone.
- Proof makes rewards fair.
- A ledger makes proof durable.
- Durable proof unlocks credible sponsorship.
VerifAI establishes that the grow was real. The ledger ensures the record can't be rewritten later — enabling public auditability and partner confidence.