While DePIN projects, in theory, attempt to provide real utility to crypto, there are few that truly solve real-life problems, have a sensible business model capable of disrupting existing companies and cannot be easily spoofed. Most are simply solutions in search of a problem. One notable exception is a flight-tracking network called Wingbits. Why? Because it addresses a Web2 problem by solving it with Web3 incentives. For anyone who has ever tracked a flight such as BA117 from London to New York, you may have used websites like FlightAware or Flightradar.
STORY CONTINUES BELOW
Figure 1: Wingbits flight tracking map
Source: Wingbits – Transforming Flight Tracking.
Flight-tracking companies generate millions in revenue by selling flight data to aviation companies and to buyers like financial analysts who monitor private jet movements for mergers and acquisitions. These companies also earn revenue from ads and subscriptions on their platforms. However, their capital expenditure does not include significant infrastructure and hardware expenses. This is because the aviation surveillance technology, called ADS-B receivers, is a hardware which requires antennas and Raspberry Pis, purchased and configured by aviation enthusiasts. These enthusiasts expect little in return, often receiving just a free subscription to their favourite flight-tracking platform.
The main problem is that enthusiasts are not incentivized to maximize the quality of data for these networks. Without marginal incentives, ADS-B receivers are often poorly placed — for instance, in lounge room corners or oversupplied in densely populated urban areas, leading to weak coverage in rural regions.
Figure 2: (LHS) Traditional ADS-B receiver, (RHS) Wingbits miner
Source: Wingbits – Transforming Flight Tracking.
Wingbits is revolutionizing flight tracking by incentivizing enthusiasts to set up stations strategically, based on altitude, while utilizing a system similar to Uber’s hexagonal hierarchical spatial index. This approach ensures optimized coverage, higher-quality data and most importantly, fair rewards for contributors to the network. They achieved coverage of 75% of the largest networks with only 1/11th the number of Wingbits stations. This high level of efficiency, combined with an expected rollout of 4,000+ stations, is anticipated to surpass traditional flight-tracking networks by a significant margin, delivering better-quality data to end customers.
The next family dinner conversation explaining this concept will come easily as we can now point to a real-world use case, driven by crypto incentives, that everyday people can understand.
Similar to market cycles, the demand for compute experiences peaks and troughs. GPUs can be expensive, and supply constraints make them even more so. Unlocking idle compute on consumer devices is not a new concept, but solving the synchronization challenge across multiple devices is. Exo Labs is a pioneering project achieving breakthroughs in edge computing, enabling users to run models on everyday consumer-grade devices, such as household MacBooks. This means sensitive data remains under your control, reducing risks associated with cloud-based storage or processing.
Figure 3: A 9-layer model is divided into 3 shards, each running on a separate device
Source: Transparent Benchmarks – 12 Days of EXO, EXO Labs.
Exo Labs has developed a novel software infrastructure called pipeline parallel inference, which enables a large language model (LLM) to be split into “shards,” allowing different devices to run separate parts of the model while remaining connected over the same network. This approach offers various advantages such as reduced latency, enhanced security, cost efficiency and most importantly, privacy benefits.
Exploring privacy further reveals Bagel AI, a project that has developed ZKLoRA (Zero-Knowledge Low-Rank Adaptation), a privacy-preserving approach to fine-tuning LLMs. This innovation enables the creation of specialized models for industries like legal services, healthcare and finance, allowing sensitive data to be used for reinforcement learning without risking confidential information leaks.
While privacy preservation is a hot topic, a bigger challenge for most LLMs is the hallucination problem, a response generated by AI that contains false or misleading information presented as fact. A portfolio manager once told me, “Wisdom lies in synthesizing competing viewpoints to uncover the nuanced truth between two extremes.” Blocksense is a project that has developed a proprietary approach called zkSchellingCoin consensus. This method aims to overlay subjective truths from multiple sources – say, different LLMs – to arrive at a single, common truth. For example, imagine running the same query across ChatGPT, Claude, Grok and Llama. If one model provides an incorrect output, it is statistically unlikely that all four models will generate the same false result when compared against each other.
Figure 4: Overview of the zkSchellingCoin Consensus
Source: Blocksense Network – The zk Rollup for Programmable Oracles.
The zkSchellingCoin consensus could also be applied to adding verifiability to AI inference. For instance, how can we confirm that an AI agent correctly bridged USDC into the highest-yielding vault at the time of execution? Trust in AI would be significantly strengthened with an additional verification layer. If we can solve this without compromising cost or latency, it could lead to a major breakthrough in real-world use cases.
The journey from hype to reality in DePIN and AI shows that genuine innovation lies in solving real-world problems with practical and efficient solutions. Projects like Wingbits and Exo Labs prove how blockchain and AI can create meaningful impact — whether by revolutionizing flight tracking with strategic incentives or unlocking the power of consumer devices for secure and cost-effective computing. With advancements like ZKLoRA for privacy-preserving AI and zkSchellingCoin for verifiable truth, these emerging technologies are poised to address critical challenges, paving the way for a more decentralized, efficient and trust-verified future.
