Eigendrum Hit Hacker News, but Its Best Trick Is Exposing Imperfect Physics

Eigendrum reached Hacker News with 173 points and 45 comments after turning an abstract mathematics problem into something anyone can strike, hear, and question.

Created by Basel Ashraf, the browser experiment treats a drawn outline as an ideal drumhead fixed around its edge. It calculates that shape’s vibration modes, converts them into sound, and animates the resulting motion.

The conflict appears as soon as users press the circle preset. The mathematics can be correct while the synthesized result still sounds unlike the drum in your head.

That gap is what makes Eigendrum more than a polished science toy. It reveals the boundary between solving an equation and recreating a physical instrument.

The Hacker News Demo Turns Any Outline Into a Drum

Eigendrum replaces a prerecorded instrument with a numerical problem solved inside the browser.

A visitor can select one of 11 built-in forms or draw a closed outline. The available presets include familiar geometry, hand-drawn forms, and two historically important isospectral drums.

The application first turns that boundary into a triangular mesh. A mesh divides a continuous surface into small elements that a computer can analyze numerically.

It then solves for the lowest 16 eigenmodes. An eigenmode is a natural standing-wave pattern that the surface can sustain without changing its basic spatial form.

Each eigenmode has a corresponding eigenvalue. In this model, the audible frequency is proportional to the square root of that eigenvalue.

The result is an interactive virtual membrane, not a conventional sample player. Changing the outline changes the ratios among its resonant frequencies.

Users can strike different positions by clicking or tapping the surface. That location matters because each vibration mode moves differently across the drumhead.

A strike near an active region can strongly excite a mode. A strike on one of its nodal lines, where that mode remains still, contributes almost nothing.

Eigendrum visualizes these patterns as moving surfaces crossed by pale, stationary channels. Those channels resemble the nodal patterns associated with historical Chladni figures.

The interface also lets users isolate individual modes. A physical mallet normally excites several modes together, so this control favors explanation over strict realism.

Users can alter pitch, decay, brightness, and mallet width. However, they cannot manually rearrange the frequency ratios determined by the outline.

That division is central to the experiment. Shape governs the spectrum, while material-like qualities remain adjustable modeling choices.

A shape can also be entered as a polar or parametric equation. This permits repeatable forms that would be difficult to draw accurately by hand.

The URL fragment stores the selected or drawn shape. That means a particular virtual drum can travel in a link without requiring a server-side account.

The project can also export a strike as a WAV file and save an image of the plate. These features turn a momentary simulation into a shareable artifact.

That sharing model helped Eigendrum spread beyond its original deployment. The Hacker News post linked directly to a circle preset, giving readers a familiar reference before they created stranger forms.

Yet the circle also became an immediate test. Users already possess expectations about how a round drum should sound, even if the simulation models an ideal membrane instead.

Why the Circle Sounds Stranger Than Expected

The circle preset exposes a mismatch between mathematical accuracy and acoustic familiarity.

A real drum is a coupled physical system. Its membrane, shell, surrounding air, room, mounting hardware, and striking tool all influence the sound reaching a listener.

Eigendrum deliberately narrows that system. It models an idealized membrane clamped around a boundary, then synthesizes its calculated modes as decaying sinusoidal components.

A sinusoid is a pure oscillation containing one frequency. Real instruments produce more complicated interactions, including nonlinear behavior, sound radiation, and resonances outside the membrane.

The project’s open-source implementation explains which properties come from geometry and which are approximations. That transparency matters when listeners challenge the output.

Shape determines the mode patterns and their relative frequencies. It also determines which modes respond strongly at a particular strike location.

Absolute pitch does not come from shape alone. It depends on wave speed, which reflects membrane tension and mass density in the underlying physical equation.

Eigendrum exposes a pitch reference because the outline cannot supply those missing material properties. It similarly lets users adjust how quickly the sound fades.

The application uses Rayleigh damping, a proportional model that makes higher frequencies decay faster. This helps the output behave more like a struck object than a sustained chord.

Mallet width changes how localized the impact becomes. A broader impact distributes force across more of the surface and generally suppresses very fine spatial modes.

Contact time also affects the upper spectrum. A physical strike lasts for a finite interval, limiting its ability to energize modes with much shorter periods.

Those decisions can improve plausibility, but they do not produce a complete acoustic drum. The synthesized sound remains an interpretation of membrane behavior.

That distinction surfaced in public feedback before the Hacker News appearance. One Reddit commenter said the circular example sounded distorted and suggested that the mathematical model was missing something.

Ashraf agreed that the result was imperfect and said he was working on it. He later edited the response to say the issue had been fixed.

Other feedback was less concerned with realism. Users praised the concept, shared unusual shapes, requested keyboard shortcuts, and asked for easier shape sharing.

Ashraf added number-key controls for modes after one request. He also pointed users toward copying the URL when they wanted to exchange a specific shape.

This rapid feedback cycle illustrates the advantage of an open browser experiment. A technical criticism does not remain confined to a paper or repository issue.

Users can challenge the result with a familiar preset, then immediately compare the next revision. The disagreement becomes part of the development process.

Still, a corrected audio bug does not remove the broader limitation. An ideal membrane should not be judged as a complete model of an acoustic instrument.

The better question is whether the simulation faithfully presents the assumptions it actually makes. On that standard, Eigendrum provides unusually direct evidence.

Eigendrum Explains How Shape Becomes Sound

The project’s main achievement is making a chain of numerical methods feel like one immediate gesture.

When a user finishes an outline, Eigendrum overlays a lattice of right-isosceles triangles. It retains the triangles that fall inside the boundary and adjusts the edge toward the original curve.

The application then repairs awkward elements and smooths interior points. This preprocessing matters because a distorted mesh can contaminate every later calculation.

Next, the program builds stiffness and mass matrices using linear finite elements. The finite element method approximates a continuous physical problem across many small connected regions.

For a membrane fixed at its edge, the target equation is a Laplacian eigenvalue problem with Dirichlet boundary conditions. Dirichlet conditions keep displacement at zero along the rim.

The solver searches for the lowest eigenvalue and eigenfunction pairs. These pairs represent the first natural frequencies and their corresponding vibration patterns.

Eigendrum uses block inverse iteration with a Rayleigh-Ritz projection. The approach concentrates computation on the lower spectrum, which dominates the audible result.

According to the project documentation, a drum with approximately 2,000 unknown values solves in about 700 milliseconds inside a browser worker.

A worker runs expensive JavaScript away from the main interface thread. This lets the visualization remain responsive while the numerical solver completes its work.

Once the modes are available, the application projects the virtual mallet onto each spatial pattern. That projection estimates how strongly the strike excites each mode.

The mode amplitudes also account for normalization and frequency. An impulsive force initially changes velocity, giving the resulting response a factor inversely proportional to frequency.

The application then combines decaying sinusoids for playback. This final stage translates mathematical modes into something a browser can send to its audio system.

The process is noteworthy because it has no application backend. Meshing, solving, animation, synthesis, and export happen on the user’s device.

The repository also has no runtime package dependencies or build requirement. A local copy can run from its static files, while automated tests use development tooling separately.

That architecture makes the Eigendrum drum simulator easy to inspect. It also limits the privacy exposure associated with sending custom shapes to a remote computation service.

The project says URL fragments carry shared shapes without being sent to the server in normal web requests. Analytics is configured to avoid recording fragment contents.

Formula input receives additional defensive treatment. Expressions pass through a custom parser instead of JavaScript evaluation, reducing the risk of a shared formula becoming executable code.

The formula system rejects curves that fail to produce a value, intersect themselves, or become too thin for an honest mesh.

That last refusal is important. Numerical software can always display a number, but a number produced by an inadequate discretization does not deserve confidence.

By declining shapes it cannot mesh reliably, Eigendrum communicates a rare engineering principle: visible output is not the same as a valid result.

The Real Opponent Is the Black-Box Science Demo

Eigendrum pressures educational software that hides its approximations behind an attractive animation.

Many interactive science pages present a simulation without showing how its result was tested. Users must trust that the motion and sound correspond to the advertised equation.

Eigendrum takes the opposite route. Its repository includes implementation details, a test suite, accuracy benchmarks, and known spectra for comparison.

The benchmark checks shapes whose eigenvalues have analytical solutions. These include a square, rectangle, right triangle, and disk.

For the lowest eight modes, the documented worst relative errors decline as the mesh grows. At 6,000 nodes, the listed errors range from 0.160% to 0.233%.

The convergence pattern closely follows the second-order behavior expected from the chosen finite elements. This does not validate every modeling decision, but it tests the numerical core.

The documentation also highlights a useful sign check. Conforming finite element estimates should approach these eigenvalues from above, rather than undershooting them.

A value below the exact answer would therefore indicate a defect instead of ordinary mesh coarseness. The test suite reportedly asserts that this does not occur.

Another check examines strike position on a circular membrane. A center strike strongly excites the radially symmetric fundamental while leaving modes with central nodal lines nearly silent.

That result was not inserted as a special audio rule. It emerges from projecting the strike onto the calculated spatial modes.

This evidence changes the nature of criticism. A listener can still dislike the sound, but that reaction does not automatically identify an eigenvalue error.

The disagreement must move downstream. It may concern damping, mallet behavior, modal balance, audio gain, or omitted physical structures.

That is a healthier argument than treating one generated sound as proof of the entire model. It separates the solved mathematics from the chosen presentation.

The approach also creates a practical teaching path. A student can hear a mode, locate its nodal lines, change the strike point, and observe the resulting mixture.

The experiment therefore joins a broader class of explorable explanations. These tools let users manipulate a model while keeping the causal relationship visible.

Unlike a static diagram, an interactive model can expose its own failure cases. An awkward outline, coarse mesh, or surprising tone invites investigation.

This is also where personal knowledge management becomes relevant for technical learners. Capturing observations across shapes can turn casual experimentation into a reproducible study record.

Eigendrum does not automatically create that record. Its contribution is making each variable concrete enough that a learner knows what evidence to capture.

The Hacker News response reflects that appeal. The project offered an unusual combination of immediate play, inspectable code, and a famous unresolved-then-resolved mathematical story.

Two Different Drums Can Share the Same Spectrum

Eigendrum’s deepest feature demonstrates that sound frequencies do not always reveal a drum’s complete shape.

Mathematician Mark Kac popularized the question “Can one hear the shape of a drum?” in a landmark 1966 paper.

The question asks whether every resonant frequency of an ideal drum uniquely determines its boundary. In mathematical terms, does a Laplacian spectrum identify the domain?

The eventual answer was no. Carolyn Gordon, David Webb, and Scott Wolpert constructed different planar domains with identical spectra.

Their 1992 result supplied nonisometric, simply connected examples. Nonisometric means one shape cannot be transformed into the other while preserving all distances.

Eigendrum includes a pair based on this construction as Kac drum I and Kac drum II. Each uses seven congruent right-isosceles triangles arranged differently.

One resembles a hook, while the other resembles an arrow. They share area and perimeter, but their more striking property is a matching set of eigenvalues.

The application calculates both spectra and draws them on the same scale. Users can switch between the two shapes and compare the corresponding frequencies.

The project reports matching values through the first 12 displayed modes. Its documentation explains why the chosen mesh also preserves the relevant edge geometry.

This is where “hear shape of drum” stops being a catchy prompt and becomes a challenge to intuition.

Most users expect an object’s sound to function as an acoustic fingerprint. That expectation often works practically, but it fails as a universal mathematical claim.

A spectrum describes the allowed frequencies, yet it does not preserve every piece of spatial information. Two boundaries can encode those frequencies in different geometric arrangements.

Strike position adds another layer. Isospectral drums share eigenvalues, but their eigenfunctions live across different shapes.

A localized strike interacts with those mode shapes, not merely their frequencies. Therefore, a particular impact can create different mixtures even when the available spectra match.

This does not reverse the theorem. It shows that hearing a single strike includes information about excitation, while Kac’s question concerns the full frequency spectrum.

The distinction is subtle enough to disappear in a textbook summary. Eigendrum makes it harder to miss because users can move the strike themselves.

The feature also guards against a shallow interpretation of the project. Eigendrum is not claiming that every outline produces a uniquely identifiable tone.

Instead, it visualizes both sides of the spectral geometry story. Geometry shapes resonance, but resonance does not always reconstruct geometry.

That intellectual reversal gives the project lasting value after its front-page attention fades. It invites experimentation while preserving the theorem that limits the experiment’s premise.

What to Watch After the Hacker News Attention

Eigendrum’s next test is whether community feedback improves its acoustic model without hiding where approximation begins.

The first signal is audio validation against real measurements. Current analytical benchmarks test the eigenvalue solver, but listeners are reacting to the final synthesized strike.

Recorded membranes with documented dimensions, tension, density, and impact locations would provide a stronger end-to-end comparison.

Such testing could separate numerical frequency error from damping, amplitude, and radiation differences. It would also clarify which controls deserve more physically meaningful ranges.

The second signal is performance across unusual user-created shapes. Built-in forms are useful, but arbitrary outlines expose weak meshing and numerical conditioning quickly.

The project already rejects self-intersections and shapes too thin to mesh honestly. Future bug reports will show whether those safeguards cover common edge cases.

Performance on mobile hardware also matters. A browser worker prevents the interface from freezing, yet limited processors can still make dense solves feel slow.

The third signal is whether the project develops into a learning platform or remains a focused experiment.

Community requests have already produced keyboard controls and sharing discussions. Additional explanations, comparison views, and reproducible experiments could deepen educational use.

However, feature growth carries a risk. Too many controls could blur the clean separation among shape, material, strike, and synthesis choices.

Preserving that separation is more important than making every result sound pleasing. A scientifically honest tool should not quietly tune geometry-derived frequencies toward familiar musical intervals.

Eigendrum could also support structured exports of eigenvalues, mode data, or mesh information. That would let students and researchers compare results outside the interface.

Its MIT license and static architecture lower the barrier for such extensions. The repository showed 106 stars and two forks when reviewed on August 16, 2026.

Those counts remain modest compared with large developer projects. Still, they indicate that the Hacker News exposure translated into attention for the implementation, not only the demo.

The decisive question is whether contributors engage with the mathematics, acoustics, accessibility, and tests together. Improvements in only one layer can weaken the experience elsewhere.

Try the circle, strike its center, then move toward the edge. Isolate several modes before drawing an asymmetric form and repeating the experiment.

Do not ask only whether it sounds like a commercial drum. Ask which differences belong to shape, which belong to the strike, and which come from the model.

That shift from passive listening to inspectable questioning is why Eigendrum deserves attention beyond one Hacker News cycle.