Why there's no speed number on this website

Lamplit exists because culling is slow. That is the entire pitch. And there is not one performance figure anywhere on this site — no multiple, no frames-per-second, no “loads 500 RAWs in n seconds.”

That is deliberate, and it is worth explaining, because the absence is more informative than a number would be.

What we have

There is a benchmark harness in the codebase. It exists, it runs, and it was written before any of this marketing was. What there isn’t is a committed set of results from it on known hardware with a written-down method.

The performance targets that appear in the project’s own internal documents are labelled, in those documents, as design targets, not measurements. They are what the architecture was built to hit. Whether it hits them on your machine, with your cards, is a different question and one we have not answered in public.

So: a harness, yes. Numbers we would stake the brand on, not yet.

Why we won’t round up in the meantime

Photographers talk to each other. Constantly. You hire each other as second shooters, you compare notes at every wedding, you post in the same three forums. A category this small has no room for a claim that falls apart the first time someone tests it on a Sunday afternoon.

And culling benchmarks fall apart easily. A number for “500 RAWs” means almost nothing without saying which RAWs — a 1.7MP embedded preview from an a6000 and a 24MP full-resolution preview from a GR III are not the same work, and the difference is a property of the camera, not of the software. Nor without saying whether the cache was cold. Nor what the disk was. Nor whether the thumbnails had been warmed by a previous run, which is the single easiest way to publish a figure that is technically true and completely misleading.

Every one of those is a lever you can pull to make your own software look fast. We would rather not be in the position of having pulled one.

What we’ll publish instead

When there are numbers, they will arrive as a page of their own, and it will contain, at minimum:

• The exact hardware. Both platforms, including a mid-range Windows laptop rather than only the fastest machine in the room.
• The exact files: which camera, which format, what preview size, how many.
• Cold cache and warm cache, separately and labelled.
• What the comparison tool was doing, if there is one, and its version.
• Enough method that you can run it yourself and tell us we’re wrong.

Until all of that exists, this site talks about speed qualitatively and leans on the numbers we can stand behind: 17 RAW formats that work, verified end-to-end against 40 real camera files, on 2 platforms. Those are on the formats page, with the camera bodies named.

The part that isn’t a technicality

There is a version of this post that is a marketing move — the humble-brag where withholding the number implies the number is enormous.

That is not what this is. We do not know what the number is on your hardware. We think it is good, the architecture was built around making it good, and the people using the beta have not been shy about the parts that aren’t. But “we think” is not a benchmark, and printing it as one would make every other claim on this site worth less.

If speed is the reason you would switch — and it should be — then the honest answer today is: run it on your own cards during the beta and decide for yourself. That is the only benchmark that was ever going to matter to you anyway.

Get early access when a wave opens, and bring the biggest card you own.