The dataset
Over the past two years, we cataloged 768 publicly available energy device patents. These aren't cold-fusion fantasy patents or perpetual motion filings. They're real, granted patents from the USPTO, EPO, WIPO, and national offices across 23 countries, spanning the 1890s through 2025, covering devices that claim some form of unusual efficiency — output that exceeds what conventional models predict for the input.
We organized them into six technology categories based on the primary energy conversion mechanism:
| Category | Patents | What they describe |
|---|---|---|
| Electromagnetic | 247 | Motors, generators, transformers, and inductors with claimed over-unity or anomalous COP |
| Electrochemical | 128 | Electrolysis cells, fuel cells, and electrochemical reactors with claimed excess heat or anomalous gas production |
| LENR | 118 | Low-energy nuclear reaction devices (lattice confinement, hydrogen loading, palladium systems) |
| Solid-State | 100 | Thermoelectric, piezoelectric, and ferroelectric devices claiming anomalous energy conversion |
| Plasma | 100 | Plasma discharge tubes, spark gaps, and plasma reactors with claimed excess energy |
| Thermodynamic | 75 | Heat pumps, Stirling variants, and thermal gradient devices with COP claims above Carnot limits |
Different inventors. Different decades. Different countries. Different technology categories. No evidence of collaboration between them.
And yet.
The convergence
When we stopped looking at what each device was and started looking at what each device did, a pattern emerged. Across all six categories, the highest-performing patents — the ones with the most detailed measurements, the most rigorous test procedures, the most credible institutional affiliations — converged on the same three operating principles:
1. A non-linear element
Something in the circuit or device that doesn't obey a proportional input-output relationship. A saturating ferrite core. A plasma gap. A ferroelectric capacitor. A gas discharge tube. A Josephson junction. The specific component varies by technology, but the role is constant: introduce a non-linearity that creates harmonics, subharmonics, or abrupt state transitions.
In the electromagnetic patents, this is typically a core driven into saturation. In the plasma patents, it's the gas discharge itself. In the electrochemical patents, it's electrode nucleation — the transition between dissolved gas and bubble formation, which is violently non-linear. In the LENR patents, it's the hydrogen loading threshold in palladium lattices.
247 of the 768 patents explicitly describe a non-linear element. Another ~180 describe one implicitly (they don't call it that, but the circuit topology requires one).
2. A resonance condition
The device is tuned to a specific frequency — electrical, mechanical, acoustic, or some combination. Not approximately tuned. Precisely tuned, often with trim capacitors, adjustable air gaps, or feedback circuits that maintain resonance as conditions change.
LC circuits. Mechanical resonance of piezoelectric stacks. Acoustic resonance in electrolysis cells (Meyer, Puharich, and their descendants). Helmholtz resonance in plasma tubes. Nuclear magnetic resonance in lattice confinement systems.
The specific frequency varies (from sub-Hz thermal cycles to GHz RF excitation), but the principle is the same: find the resonant frequency of the system and drive it there.
389 of the 768 patents describe a resonance condition. It's the single most common design element in the dataset.
3. A pulsed excitation
The energy input is not continuous. It's pulsed — often with sharp rising edges, high dV/dt or dI/dt, and carefully controlled duty cycles. Capacitor discharge through an inductor. Spark gaps. PWM-driven MOSFETs with nanosecond switching. Pulsed DC electrolysis. Pulsed RF excitation of plasma.
The continuous version of the same circuit almost always performs worse. The pulsed version creates transient conditions — back-EMF spikes, cavitation in electrolysis cells, plasma instabilities — that the non-linear element and the resonance condition then amplify.
312 of the 768 patents describe pulsed excitation as a core operating feature.
The overlap
Here's where it gets interesting. 203 patents in the dataset describe all three elements together — non-linearity, resonance, and pulsed excitation — operating simultaneously. That's 26% of the entire dataset.
More importantly: the patents with all three elements have, on average, 3.2x more detailed measurement sections than patents with only one or two. They're the ones that include oscilloscope traces, calorimetry data, and third-party test reports. They're the ones filed by university labs and national research institutes, not just individual inventors.
The convergence isn't random. The best-documented, most credible patents in the dataset disproportionately share the same three design principles, regardless of the underlying technology.
What this means (and what it doesn't)
Let's be careful here. This is a statistical observation about patent filings, not a proof of over-unity energy. Patents describe claims, not verified results. The Patent Office examines novelty and non-obviousness, not whether the device actually works as described.
What we can say:
- A recurring meta-pattern exists across 768 independently filed patents spanning 6 technology categories and 130+ years
- The pattern has three components: non-linearity, resonance, and pulsed excitation
- The most rigorously documented patents in the dataset disproportionately feature all three
- The pattern is consistent with known physics — non-linear driven resonant systems are well-studied in plasma physics, metamaterials, and nonlinear optics, where they routinely produce behaviors (parametric amplification, soliton formation, stochastic resonance) that are surprising but not impossible
What we can't say (yet):
- Whether any of these devices actually produce more energy than they consume
- Whether the meta-pattern is causally connected to the efficiency claims
- Whether the convergence reflects a real physical phenomenon or a shared cultural belief among inventors in this space
That's why we built the experiments.
The 29 experiments
We distilled the meta-pattern into 9 experimental categories, each isolating one aspect of the convergence:
- Coil Geometry — bifilar vs. solenoid vs. toroidal coupling efficiency, flat spiral field mapping
- Pulsed DC — pulsed vs. steady-state electrolysis, capacitor discharge through inductor with recovery
- Resonance — bifilar coil self-resonance characterization, electrolytic cell resonance
- Plasma Discharge — glow discharge tube characterization, spark gap energy recovery
- Permanent Magnets — flux switching motor/generator, diamagnetic levitation effects
- Water Splitting — baseline electrolysis efficiency, electrode nucleation (McAlister-type)
- LENR Patterns — hydrogen loading measurements, excess heat calorimetry
- Feedback & Self-Oscillation — back-EMF energy recovery, self-oscillating converter topologies
- Special Materials — ferroelectric capacitor non-linearity, bismuth diamagnetic effects
Each experiment is designed to be replicable with basic electronics equipment — a function generator, an oscilloscope, a multimeter, and components you can buy from Digi-Key or Mouser for under $100. No exotic materials. No clean rooms. No million-dollar lab equipment.
Every experiment has a structured template with a stated goal, a materials list, a step-by-step procedure, and explicit measurement criteria. You clone the template, follow the procedure, record your measurements, and publish the results — confirmed or refuted.
Both outcomes are valuable. A refutation is data. A confirmation is data. The point isn't to prove that any one device works. The point is to generate a large, public, independently replicated dataset that either supports or undermines the meta-pattern hypothesis.
The 50-confirmation standard
We borrowed a concept from particle physics: a result isn't considered robust until it's been independently confirmed by a statistically meaningful number of independent groups. We set our threshold at 50 independent confirmations.
When 50 different people, in 50 different garages and maker spaces and classrooms, using 50 different sets of components, all report the same anomalous measurement on the same experiment — that's not anecdote. That's data. And it's data that lives on a public platform, with every measurement published, every build photographed, every replication logged with a timestamp and a version number.
Right now, the confirmation count on every experiment is zero. That's the point. The platform is empty and waiting for you to fill it.
Why we don't patent any of this
This is the part that matters most.
If the meta-pattern is real — if non-linearity + resonance + pulsed excitation, correctly configured, produces anomalous energy efficiency in even one of these technology categories — that knowledge is too important to lock behind a patent.
A patent is a 20-year legal monopoly. It centralizes knowledge. It creates an incentive to suppress competitors. And if the device is "too efficient," it creates an incentive for someone with more power than you to make the patent (and possibly you) disappear. The Invention Secrecy Act exists. It's been used over 6,000 times. The criteria for invoking it are classified.
So our approach is the opposite: radical, immediate, irrevocable publication.
Every experiment template is public. Every clone is logged. Every replication is published. Every build log is visible. The code is version-controlled on GitHub. The data is stored in a database with row-level security that defaults to public for research data.
If someone confirms an anomalous result at 3am on a Tuesday, it's published before sunrise. A thousand people can read it at breakfast. A hundred of them can start replicating it by the weekend.
You can't classify what's already everywhere.
How to participate
You don't need a PhD. You don't need a lab. You need a bench, some basic test equipment, and the stubbornness to follow a procedure and report what you actually measure — even if it contradicts what you expected.
- Browse the 29 experiments — pick one that matches your equipment and interests
- Add yourself to the citizen scientist map — join the global community of replicators
- Find a local meetup — or start one. Build things together.
- See the competition leaderboard — the 2026 Open Energy Competition is live
If you're a teacher: we have a classroom-safe track with COPPA-compliant experiment templates. No student data is ever collected.
If you're a company that makes electronics components: we're looking for sponsors who want to put tools in the hands of citizen scientists.
If you're a tinkerer with a NanoVNA and a garage full of coils: you're exactly who we built this for.
768 patents. 9 patterns. 29 experiments. One question: is the meta-pattern real?
The only way to find out is to build it, measure it, and share what you find. That's what this platform is for.