The energy systems of the 21st century face a puzzle: how to manage increasing complexity as renewables proliferate and devices grow smarter. The electrical grid is evolving from a centralized machine into a continent-spanning organism — dynamic, decentralized, and hard to control.

Smart homes, electric vehicles, thermostats, and appliances are beginning to operate autonomously. But under the surface, most of them remain tightly dependent on centralized services: proprietary, revenue-oriented cloud APIs and platform ecosystems. These platforms are designed to maximize profits and frequently optimize for lock-in, usage fees, and data monetization. In a world of growing geopolitical tension, even access to essential services can become a point of leverage [0].

What if coordination logic didn’t depend on opaque cloud services? What if our devices could participate in shared decision-making — operating together, securely and efficiently, without a central overseer? This is the motivation behind Project Hummingbird, a small but meaningful demonstration of decentralized automation using the Flow blockchain.

1. The Grid Is Changing

The rapid adoption of renewable energy sources is reshaping the demands on electrical grids worldwide. The European Union is embarking on massive infrastructure investments to prepare its power grid for a future where renewable sources (already contributing 47% of electricity in 2024) provide the large majority of energy supply [1, 2].

Simultaneously, new EU policies and R&D initiatives are guiding industrial energy producers, consumers, and manufacturers of household devices toward open, interoperable frameworks for rebalancing power supply and demand. Such adjustments typically occur in 15-minute intervals, forming the technical basis for a continent-spanning smart grid, where swarms of devices dynamically coordinate to balance the power system in the future [3].

To meet these goals, smart devices must do more than turn on and off — they must shift their time of operation and respond predictively to grid conditions. And the coordination logic that governs them needs to be transparent, robust, and resilient.

Traditionally, coordination relies on centralized service offerings — typically hosted in proprietary cloud platforms. These systems are effective in some respects but raise deep concerns about:

• Digital sovereignty
• Relinquishing personal data to service providers
• Single points of failure in infrastructure that increasingly governs heating, transportation, and basic household functioning

2. Why We Used Flow

Blockchains are, in theory, open-access, fault-tolerant, tamper-proof compute platforms. In practice, however, most are poorly suited for large scale control applications.

Flow is different. It was designed from the ground up to support scalable, composable computation. Here’s why we believe Flow is uniquely positioned to enable decentralized automation:

• Scalable, open-access computation: Flow is architected to provide horizontal scalability, targeting millions of transactions per second and on-chain data ingress on the order of gigabytes per second [4]. Anyone can deploy or interact with smart contracts. No gatekeepers can arbitrarily deny service.
• Safety-first programming model: Flow uses a resource-oriented programming language (Cadence) with native language primitives for fine-grained access control and safety by default [5]. This is especially important for real-world applications that interact with physical systems.
• Modular, easily composable resources and smart contracts: By partitioning logic into smaller building blocks with clear boundaries, easier to reason about correctness, re-use code and compose larger systems.
• Lightweight cryptographic proofs: Flow supports cryptographic event and state proofs, allowing devices with small computational capabilities (like microcontrollers) to verify correctness without trusting third-party nodes [7].
• Foundation for privacy-preserving orchestration: Flow’s computational scalability opens the door for secure multiparty computation (MPC) frameworks running on-chain [4]. These frameworks, combined with on-chain orchestration, could allow for end-to-end encrypted control logic, where devices can collaborate securely without leaking sensitive information.

3. Project Hummingbird: Prototyping Decentralized Automation

To make this vision concrete, we built Project Hummingbird: a blockchain-controlled heating system for a hummingbird feeder in coastal British Columbia.

Wild hummingbirds thrive in the mild climate of this region and are common, year-round residents. In cities like Vancouver, they rely on sugar-water feeders during rare cold snaps. To keep the nectar from freezing, people often use heat strips or lamps — but most setups are either wasteful (always on) or unreliable (we forget to turn them on).

So we asked: could a blockchain smart contract manage that heating automatically — turning it on only when temperatures drop?

The result was a complete end-to-end system:

• A Cadence smart contract on Flow defines the control logic
• An oracle feeds temperature data into the chain
• At the heart of our IoT device is a microcontroller, which listens for blockchain events and flips a switch — powering the feeder heater only when it’s actually needed

You can watch the demo video here:

From a transaction submitted in Sweden to a device responding in Canada — this end-to-end demo illustrates how decentralized logic can orchestrate a swarm of real-world devices. In my video, I have a swarm of one device; a follow-up demo with multiple smart devices across the globe is in the works.

Big thanks to Janez Podhostnik and Jan Bernatik for building this with me! Our project is open source and documents both the system and hardware details. But the goal is not to spark a wave of blockchain-powered bird feeders. Rather, it’s to show how Flow can be used as secure platform for automation, IoT, and smart home devices — using tools already available today.

4. How It Works: From Transaction to Action

The devices listen for high-level signals and respond accordingly, but they retain local autonomy. This model is fundamentally different from streaming raw sensor data to the cloud and running centralized analysis. It is lightweight, resilient, and privacy-preserving by design.

A good mental model is to think of the blockchain as the conductor of an orchestra. Each connected IoT device — like our feeder — is an individual musician: capable, autonomous, and specialized on a specific task. But just as a single musician may struggle to catch the nuances of instruments at the other end of the orchestra, it is challenging for individual machines to obtain a holistic picture of the macroscopic world state. The conductor — in our case, the on-chain control algorithm — provides the overarching coordination, turning a group of uncoordinated specialists into a collaboratively working swarm.

Let’s unpack how the system works in practice.

Our smart hummingbird feeder connects to a Flow Access Node. Via WebSockets, it streams the specific events emitted by the on-chain control logic — in our case instructions to activate or deactivate the heating. When new environmental data (e.g. from weather forecasts or external sensors) is submitted to the blockchain, the smart contract processes the update and decides how the feeder should respond. If our IoT device restarts, it automatically re-establishes its connection and queries the on-chain control smart contract to restore its state.

This demonstrates the feasibility of a fully decentralized, event-driven automation system:

• No need to invest resources into dedicated servers.
• No reliance on cloud functions prone to vendor lock-in.
• No resource-intensive polling loop.

Instead, utilizing Flow’s decentralized, open-access compute capabilities for the orchestration of smart IoT devices.

5. Beyond the Feeder: A Vision for Critical Infrastructure

While Project Hummingbird is a proof of concept, the implications extend far beyond. We have shown the possibility of orchestrating smart-home ecosystems, fleets of electric vehicles, solar panels, and eventually, critical infrastructure.

Flow’s architecture is uniquely suited to support such applications. With projected computational throughput far exceeding that of most blockchains [4], Flow will be able to accommodate ecosystems with billions of IoT devices. Its emphasis on safety, composability, and open access makes Flow a strong foundation for building resilient, user-controlled automation systems.

We envision a resilient, sustainable future:

• Consumer empowerment: Devices and data work for their owners, not for corporate profits. Flow reduces vendor lock-in and frees users from the need to surrender their data to proprietary clouds.
• Market innovation: An open, app-store-like marketplace facilitates a rich variety of customizable on-chain control algorithms for users to choose from — just like with mobile apps, everyone finds a solution for their individual needs.
• Energy efficiency becomes emergent, not enforced: Advanced control logic enables demand-side flexibility [6] and organic integration of renewable energies, making it the economically rational choice for users to contribute to grid stability and sustainability.

6. Scientific and Engineering Challenges Ahead

While promising, decentralized control comes with open challenges:

1. Trustless data retrieval: In our proof of concept, the microcontroller trusts the Flow Access Node to deliver accurate data. In the future, lightweight cryptographic event proofs will enable the IoT devices to cryptographically verify the integrity and completeness of the data it receives from the Access Node [7, 8].

2. Privacy-preserving orchestration: Coordinating behavior across devices without leaking sensitive data requires privacy-preserving multi-party computation [MPC]. Due to its scalability and performance, Flow is a suitable platform to shoulder the computational load of such protocols, but implementation is non-trivial.

3. Lightweight cryptography: IoT devices often have limited computational resources. Running BLS-based verification or MPC-friendly logic requires efficient libraries. Nevertheless, research results from the last years consistently demonstrate practicality [9].

4. Latency-awareness: Some applications may demand tighter timing. Flow currently provides finality in 5–12 seconds (as of 2025, with optimizations in progress), which is appropriate for most but not all use cases.

These challenges open up fertile ground for interdisciplinary collaboration — bridging formal verification, cryptographic soundness, and physical control systems.

7. Why This Matters

The world is trending toward increasingly automated and interconnected systems — in homes, in cities, in vehicles, and in infrastructure. Yet the control layers of these systems are often hidden, proprietary, and built on “security by obscurity.”

With Project Hummingbird, we’re exploring a new model: autonomous infrastructure, built on principles of openness, verifiability, and user agency. This is the moment when blockchains begin to provide utility beyond finance and speculation — not just tracking ownership, but helping optimize the real world around us.

References

[0] Heise, Criminal Court: Microsoft’s email block a wake-up call for digital sovereignty (2025)

[1] Eurostat, Electricity from renewable sources reaches 47% in 2024 (2025)

[2] Reuters, EU power grid needs trillion-dollar upgrade to avert blackouts (2025)

[3] EU Joint Research Centre, Energy smart appliances: Launch of EU Code of Conduct on interoperability (2024)

[4] Flow Core Protocol Vision (2023)

[5] Object-oriented languages are very intuitive for software engineers. On Flow, we use the term ‘resource’ for objects. In a nutshell, resources are just objects with built-in language primitives to easily manage access permissions. For readers interested in learning more, I recommend Michael Bogan’s blog The Power of Resource-Oriented Programming in Flow/Cadence: A Deep Dive (2023)

[6] smartEn, Position Paper: The contribution of demand-side flexibility to EU competitiveness and affordability (2025)

[7] Flow Data Availability Vision (2024), section Smartphone-sized light clients

[8] Flow Architecture → State and Event Proofs (2025)

[9]Various scientific publications, including Performance Evaluation of Elliptic Curve Libraries on Automotive-Grade Microcontrollers (2019); Performance Comparison Of ECC Libraries For IOT Devices (2024); Tiny keys hold big secrets: On efficiency of Pairing-Based Cryptography in IoT (2025)