My Old Team Is Still Winning

There is a specific emotional register you hit when you see your old team drop another paper and it's good — not "good for an academic paper" good, not "solid contribution to the literature" good, but actually good, the kind where you close the tab and sit with it for a second.

Janusch and the crew at USC are doing that right now in the 3D Gaussian Splatting space.

3DGS is one of those techniques that arrived and immediately made the prior five years of NeRF work feel like a very expensive detour. Same destination — photorealistic novel view synthesis — but Gaussians get you there faster, cheaper, and with real-time rendering that doesn't require a prayer to the CUDA gods every time you want to look at something from a new angle. The field has been sprinting since the original paper dropped, and USC is in the front pack.

The thing about watching your old team from the outside is that you know the texture of how that work gets made. You know the 2am Slack messages, the ablation tables that didn't survive the revision, the demo that broke the night before the deadline and got fixed with thirty minutes to spare. The paper lands clean and confident. It was not made that way.

Proud of them. Mildly unwell about it.