What shipping AEOS Labs taught me about 3D on the web

AEOS Labs came to me for a landing page. An advertising company backed by Varun Mayya and Tanmay Bhat, wanting a site that looked like nothing else in their category. The brief was simple: make it feel like the future. The build was not simple at all.

I worked the 3D with Ashwani Kumar, running Framer for structure and Spline for the interactive scene. That pairing is common advice in the Framer community now. In 2024, it was still something you had to prove worked, project by project.

The pitch versus the payload

Everyone wants a 3D hero. Nobody wants to think about what a 3D hero costs.

A Spline scene isn't an image. It's a runtime, a scene graph, and a render loop, all shipped into a browser tab that's also loading fonts, CMS data, and whatever else Framer is doing underneath it. Desktop Chrome on a wired connection forgives almost anything. An older Android phone on hotel wifi forgives nothing.

Every 3D decision on AEOS came back to that gap. Not "does this look good" — that part is easy, Spline makes almost everything look good. The real question was whether the scene still felt intentional once the device got worse and the network got slower. Spectacle is what convinces a client in the pitch meeting. Performance is what keeps the visitor on the page long enough to see any of it.

I stopped treating those as opposing goals somewhere in that project. They're sequential. You earn spectacle by surviving the boring constraints first: asset weight, load order, what renders before the 3D layer is even ready. Skip that step and the 3D isn't a feature. It's a tax the whole page pays.

Working the Framer-Spline seam

Framer and Spline weren't built for each other. They cooperate because both teams made embedding possible, not because the workflow is native.

That means the seam between them is where the real craft lives. Spline handles the object, the material, the camera path. Framer handles everything the object sits inside: the type, the layout, the scroll behavior, the moment the 3D canvas hands control back to the rest of the page. Get that handoff wrong and it's obvious immediately. The scene loads before the layout is stable, or the layout shifts once the scene mounts, and the whole thing reads as sloppy instead of expensive.

Get it right and it's invisible. Nobody compliments a clean handoff. They just don't notice anything wrong, which on a 3D-heavy site is the actual compliment.

The Google event

AEOS Labs got presented at a Google event. Sundar Pichai saw it.

I didn't build that page thinking about who might eventually look at it. You never do. You build it thinking about the deadline, the client's notes, the asset that's still too heavy. The audience shows up later, if it shows up at all, and you don't get to choose who's in the room.

What I took from that moment wasn't validation of the flashy parts. It was validation of the restraint. The scene held up in front of an audience with no patience for anything half-finished, because the boring work was already done: load order sorted, fallback states handled, nothing left to stutter under scrutiny. Spectacle gets you noticed. Discipline is what survives being noticed.

When 3D earns its place

I don't reach for 3D by default, and I didn't really reach for it by default before AEOS either. It earns its place under specific conditions.

The product or brand has to be spatial in some real sense: something with form, material, or depth that a flat mock genuinely can't communicate. The team has to accept that a 3D hero adds real engineering time, not an afternoon of polish. And the page has to fail gracefully. A visitor on a weak connection should get a slower but complete site, never a broken one.

Most landing pages don't clear that bar, and that's fine. A well-set page of type and a strong photograph will outperform a mediocre 3D scene every time. AEOS cleared the bar because the brief demanded it and the client was willing to pay for the engineering that spectacle actually requires.

That's the part people skip when they ask for "a 3D site like that one." They want the moment it gets seen by someone like Sundar Pichai. They rarely want the unglamorous engineering it takes to earn that moment in the first place.