How we test
Most gear reviews describe a sound and then copy the specifications off the box. The specifications are the manufacturer’s claim rather than anyone’s finding, and you can’t check a description.
So there’s one rule here, and the site is built so it can’t be broken:
No review describes a piece of gear nobody on this team plugged in and measured.
The build enforces it. A review that doesn’t name the exact revision tested, say how the unit reached us, cite a rig, and carry three of our own photographs and at least one measurement will not compile. There’s no override flag.
What we measure
- Round-trip latency, actually measured. Not the figure in the driver panel. A loopback cable, a click, and the real number at every buffer size — which is often not what the box says.
- How much clean gain a microphone really needs. Everyone asks whether a given interface can drive an SM7B without a booster, and the honest answer depends on the preamp’s noise floor at the gain you’d actually be using. So we measure it instead of repeating the forum answer.
- What a preamp does at the top of its range, which is where the specification sheet stops describing it.
- Connector and bus reality. Whether the USB-C box behaves any differently from the Thunderbolt one people pay twice as much for.
Every measurement is published with its method beside it, so somebody with the same gear can run it again and tell us we got it wrong. That’s what a number is for.
The bench
A latency figure from one host on one machine can’t be compared with a figure from another, and a table that quietly mixes them is worse than no table at all. So every review names the machine and the signal chain behind the number.
Where a computer helps, and where it doesn’t
We use language models for copy-editing, for turning our own measurement sheets into tables, and for pulling the testable claims out of a manual so we know what to go and test. They never originate a claim about a product. Nothing gets written about a piece of gear before somebody has had it on the bench.
The test before publishing is blunt: delete every sentence a model wrote. If the substance is still there, it was a tool. If the review goes with it, the review doesn’t go up.
How gear reaches us
Every review says it near the verdict, in plain words: bought, loaned, or supplied. Loaned units go back. Lenders don’t see the review before it’s published and get no approval over it.
We don’t sell reviews. A maker can buy bench time, and the results publish whatever they turn out to say.