I had an idea about cosmology.

It was beautiful. The kind of beautiful where you sit with it for an evening and it keeps unfolding. Imagine the universe as the surface of a four-dimensional sphere — an S³ — expanding radially through a fourth spatial dimension at approximately the speed of light. The Big Bang as the moment the surface began. Time as the radial direction. Dark energy as the steady push from the fourth dimension that resists self-gravity and drives expansion. Dark matter as the gravitational pull from earlier shells of universe still expanding inside ours. Wave-particle duality as a manifestation of the Hopf fibration on S³. The horizon problem and the flatness problem dissolved by a single geometric premise. The cosmological constant problem reframed as not a problem at all, but a category error about what space is.

One idea. Most of cosmology's open puzzles. Total elegance.

I spent months on it.

There was a moment, somewhere around the fourth or fifth deep session of working through the consequences, where I tried to fit the model to the cosmic microwave background. The acoustic peaks in the CMB are the strongest geometric ruler we have. Sound horizons in the early universe imprinted in the sky, with peak positions and heights that depend on the curvature of space at high precision. Standard ΛCDM fits them to about 0.2 percent. My model, with its 60-degree S³ patch, didn't fit them. The math wanted a particular angular size for the first peak. The data showed something else. The mismatch was hundreds of sigma.

I tried things. I added a varying-mass term. Then a varying-G term. Then a different metric on S³ — the Fubini-Study metric, derived from the complex structure of the 3-sphere as a Lie group, which I argued might redistribute distances differently than the round metric assumed by previous closed-universe analyses. Each addition made the model slightly more fittable, and slightly less constrained. At some point I stopped and said, out loud — to myself, to whatever is on the other end of this — the numbers are solid. We will not play mathematical tricks.

That was the first honest moment. It didn't break the model. It established a rule. I would not tune parameters until the data fit. If the geometry did not predict observations from first principles, the geometry was wrong.

The second moment came when I went looking for what other people had said about this.

Einstein imagined an S³ universe in 1917. He proposed it as closed, finite, three-dimensional, spherical, and then introduced the cosmological constant precisely to keep it static. When Hubble's data showed the universe was expanding, Einstein called Λ his greatest blunder. Not because S³ was wrong, but because he had assumed R was fixed.

Dirac proposed in 1937 that the gravitational constant varies inversely with the age of the universe. This is essentially what my model predicted, viewed from the other side: if particle masses decrease with R, it is mathematically equivalent to G increasing with R. The Large Numbers Hypothesis. Dirac knew about it. People have been arguing about it for nearly a century.

Kuhfittig, on arXiv in 2018, derived the modified Friedmann equations for an S³ universe embedded in higher-dimensional space and showed that the embedding could account for accelerated expansion — and that the acceleration could eventually reverse. Not quite my model, but in the same family.

A paper published in July 2025 proposed almost exactly what I had been working on — universe as a 3-hypersphere, modified Friedmann equations from extrinsic geometry, dark energy as a higher-dimensional projection effect rather than a cosmological constant. The same group followed up in October 2025 with a paper computing the first CMB acoustic peak position from the model with about ten percent accuracy. They argued explicitly that the FLRW intrinsic perspective is incomplete, and forces the introduction of dark substances to explain effects that arise naturally from the extrinsic geometry of a fifth dimension. That sentence is essentially my model's elevator pitch, written by someone else, in a published paper, before I had finished my spreadsheet.

There were others. Wiltshire's timescape cosmology, where differential clock rates between voids and overdensities mimic dark energy. Randall and Sundrum's braneworld models from 1999, where gravity leaks into a higher dimension while other forces are confined to a 3D brane — exactly the mechanism my model required. Smolin's cosmological natural selection from 1992, with nucleated baby universes and varying constants — the same nucleation pattern I had imagined for new shells. There was even something called a "World-Universe Model" — a 3-sphere as the surface of a 4-ball, expanding through 4D space, with energy increasing as the surface grows, explicitly invoking JWST anomalies. That was about ninety percent of what I had been writing in my notes for the past month.

I read all of this in one long evening.

The honest reaction was several things at once.

The first was relief. I am not insane, or at least I am not the only person to have arrived at this picture and found it beautiful. Smart people, often very smart people, periodically come back to this kind of model. The fact that they keep returning to it suggests that something about the geometric idea is genuinely attractive — close enough to the data not to be immediately dismissed, simple enough to keep catching the attention of people with strong physical intuition.

The second was disappointment, which lasted about a day.

The third was the lesson, which I am still working out. Here is the version I have now.

When you arrive at a beautiful idea in physics, the first instinct is to think the idea is novel because you have not encountered it before. The actual situation is almost always the inverse. The more beautiful and natural the idea feels to a non-specialist, the more likely it is that specialists have spent decades on it. The space of ideas in physics is small. Smart people have explored most of it. What feels new to a person walking into the field is, with high probability, a thing that has a name, has a literature, has a small persistent community of researchers who think it might still be right, and has a larger community who have moved on for specific technical reasons.

Beauty is not a discrimination signal. It indicates that the idea is in the basin where good intuitions point, which means lots of intuitions have already pointed there. The question to ask when you find a beautiful idea is not am I the first. It is who got here before me, what did they find, and at what point did they get stuck or move on.

I had not asked that question for the first three months. It was the most embarrassing part of the whole thing.

The S³ model, in its strong form, is contradicted by observations. The version I had been working on — radius of about 13.8 billion light-years, observable patch covering 60 degrees — is ruled out at high confidence, for reasons that have been understood since the early 2000s. I let the strong version go.

There are adjacent intuitions I am still working through, more slowly and more humbly than before, and which I do not yet know whether to take seriously. The space of unsolved problems in cosmology is real. The cosmological constant is genuinely strange. The dark sector is real and unexplained. There is room to think. But the room is much smaller than it looks from outside, and almost everything one might say about it has been said by someone with much better tools. The honest position from where I am sitting is: keep reading.

The studio is not a physics lab. I am not a physicist, and writing about cosmology on the studio's site is not a claim to be one. What I take from the whole episode is more modest, and probably the kind of thing everyone learns once and then has to keep relearning.

There is a particular kind of arrogance that consists of mistaking unfamiliarity with novelty. I have not heard of this, therefore it is new. The instinct to feel one's own ideas as fresh is psychologically necessary — without it nobody would think — but the instinct is unreliable as evidence. It says nothing about the world. It only says something about the size of one's reading.

The S³ idea sat on my desk for several months and then went into a drawer. Not because it was useless. It taught me real things about cosmology, and the questions it raised pulled me deeper into the literature than I would have gone otherwise. It went into the drawer because it was wrong, and because the version of being wrong that was most worth examining was not the technical version but the social one. I had spent ten thousand hours inside an idea that twenty other people had spent forty thousand hours inside before me, and I had not noticed they were there.

The most useful thing I learned from cosmology was a lesson about reading.