SkyTensor is a Bittensor subnet that creates a global, decentralized sensor network for civilian airspace monitoring. Sensor miners operate microphones, cameras, and radio receivers pointed at the sky. Processing miners run AI models to detect, classify, and track aerial objects. Validators verify quality through canary injection and Byzantine fault-tolerant fusion. The result: airspace intelligence with quantified confidence, produced by a network that evolves with the threat landscape.
Drone violence is poised to become an uncontrollable threat to civilian life worldwide. It is no longer hypothetical.
In combat zones, drones are killing civilians daily. Ukraine's experience has proven both the threat and the possibility of defense: a distributed network of 14,000 acoustic sensors (Sky Fortress, $400-1000 each) and 200,000 citizen observers (ePPO app) has closed 80% of Ukrainian airspace to Russian drone attacks 12. This works. But it required a wartime government to organize.
Outside combat zones, the threat is emerging. Criminal organizations, lone actors, and accidents involving commercial drones pose growing risks to airports, critical infrastructure, public events, and civilian populations. The global counter-drone market is projected to grow from $2.6B in 2026 to $19-69B by 2030-2035 3. The demand for airspace monitoring is real and growing.
Existing solutions are centralized and expensive. Commercial counter-drone systems (Dedrone, DroneShield) cost $50K-500K+ per installation. Military systems are classified and inaccessible to civilians. The distributed volunteer model (Flightradar24's 50,000+ ADS-B receivers 4) proves that civilians will operate sensors voluntarily — but ADS-B only tracks cooperative aircraft that broadcast their position. Adversarial drones don't broadcast.
The missing piece is decentralized, incentivized, multi-modal detection of non-cooperative aerial objects. That's what SkyTensor builds.
Sensor miners operate physical sensors (microphones, cameras, SDR receivers, smartphones) in their local airspace and upload raw captures with cryptographic attestation. A repurposed smartphone or a $100 Raspberry Pi kit is enough to start mining.
Processing miners run AI detection, classification, and tracking models on sensor data assigned to them by validators. An analysis miner in California or Greenland can analyze acoustic data from Ukraine — the global distribution of processing miners is the structural defense against local adversaries who might try to blind the network.
Validators orchestrate the system: route sensor data to processors, inject synthetic canary targets to measure accuracy, apply Byzantine fault-tolerant fusion across multiple processor reports, compute confidence levels for each geographic cell, and set weights.
Intelligence consumers — Civilians in conflict zones, airport operators, municipal governments, researchers and citizen scientists consume airspace intelligence via API.
SkyTensor uses Bittensor's multiple incentive mechanisms:
- Mechanism 0: Data Collection (~40% emissions) — Sensor miners scored on geographic coverage, uptime, data quality, and multi-modal operation
- Mechanism 1: Data Processing (~60% emissions) — Processing miners scored on detection accuracy (canary calibration), classification correctness, speed, and multi-modal fusion capability
The two mechanisms create a virtuous cycle: more sensor coverage produces more data, which attracts more processing miners, which produces better intelligence, which attracts more consumers, which funds more emissions.
The same sensor infrastructure supports multiple intelligence products beyond the core drone detection mission:
| Use Case | Description | Revenue Potential |
|---|---|---|
| Drone detection | Core mission — civilian airspace safety | Government, NGO, infrastructure contracts |
| Commercial monitoring | Airport, event, and facility airspace security | B2B API subscriptions |
| Weather phenomena | Acoustic and visual weather monitoring | Meteorological data licensing |
| Citizen science / UAP | Anomalous aerial phenomena logging | Hobbyist subscriptions, network density |
| Wildlife monitoring | Bird migration tracking, corridor monitoring | Conservation research funding |
Each scope extension is a different processing task on the same underlying sensor data. New processing miners with domain-specific models join and compete without requiring changes to the sensor network.
See DESIGN.md for the full architecture document.
This subnet is in active design. See DESIGN.md for the architecture and research/ for the industry research backing the design decisions.
Footnotes
-
Sky Fortress acoustic detection network: United24 Media. NATO expansion funding: Militarnyi. ↩
-
ePPO crowdsourced air defense app: Ukrinform; Eurasian Times. ↩
-
C-UAS market projections: Precedence Research; MarketsAndMarkets; Fortune Business Insights. ↩
-
Flightradar24 distributed network: Flightradar24. 50,000+ volunteer ADS-B receivers on commodity hardware. ↩