Blockchain carbon credits, in Mongolia
Carbon-credit systems live or die on believable measurement, not branding.
When people hear "blockchain carbon marketplace" they roll their eyes, and honestly — fair. The first generation of these systems was almost entirely vibes and venture money. The credits were unverifiable. The supply was infinite. The buyers were imaginary.
The picture is more interesting when you get close to the actual measurement problem.
What actually matters
It turns out the chain is the easy part. The hard part is MRV: monitoring, reporting, and verification.
A carbon credit is a promise that one tonne of CO₂ has been kept out of the atmosphere or pulled back into the ground. The credit is only as good as the data behind that promise. If the data is bad, the credit is worth less than zero — it's actively misleading the buyer about climate impact.
So the question isn't really "should this be on a chain?" It's:
- How do you measure the carbon being sequestered or avoided?
- How often, and how cheaply?
- How do you make the measurement believable to a buyer who's never seen the land firsthand?
If you nail those, the chain layer is a clean primitive on top — settlement, retirement, and a permanent audit trail. If you don't nail those, the chain just makes the bad data more permanent.
The Mongolia angle
Mongolia is one of the most interesting places on Earth for this. We have:
- Massive grasslands
- A coal-heavy energy mix actively transitioning to solar
- Real, measurable, visible environmental change happening at the scale of satellite pixels
That last bit is why remote sensing matters here. Satellite and drone imagery, paired with vegetation and rangeland models, can turn "is this landscape degrading or recovering?" into something measurable enough to audit.
What we got right
- Measurement first, marketplace second. The trading layer should sit on top of verified data. Not the other way around.
- Real photos, real coordinates. Every credit ties back to specific pixels on specific imagery on specific dates.
- Local registries before global. Trust gets built slowly. Better to settle locally first.
What we got wrong (the first time)
- Underestimating offline data flows. Field teams in the steppe don't have stable connectivity. The first version of our intake tool assumed they did. Painful retrofit.
- Over-trusting the model. Our first vegetation model was overconfident on edge cases. We now require human review on any flagged anomaly above a threshold.
Where this goes
The next step is scaling measurement without scaling cost. More satellites, fewer drone flights, smarter models. Less ceremony, more measurement.
If you're working on this — climate-tech, MRV, anything that touches real environmental data — drop me a line. The space is small and there's room for everyone who's serious.