Jev images in MotherDuck: Sonnet quality, 1/25th cost
Hamilton Ulmer says the image benchmark he ran scores worse than MotherDuck's hot dog bench, and that the confidence scores are what rescue it. If you are classifying images at analytics scale, the gating matters more than the headline accuracy.
Built with it
Matthew Berman gates his Meta ad loop on Jev before Claude runs
Matthew Berman's five step Meta ads loop has Jev return KILL, WATCH, SCALE or PATCH_PAGE_FIRST every 15 to 30 minutes. Zero conversions at twice target spend is a KILL, CPA at or below 0.8x with eight conversions is a SCALE. Claude runs only after Jev says SCALE, and a message match check against the landing page can block that.
don't murder winning ad hooks because your landing page can't close.
Someone measured it
Ulmer says Jev images match Sonnet at a 25th of the cost
Hamilton Ulmer says Jev image support in MotherDuck SQL reaches comparable quality to Sonnet at ten times the speed and a twenty fifth of the cost, with classification, scoring and bounding boxes. He also says the benchmark behind it scores worse than their hot dog bench, and that Jev's confidence scores can be used to drastically improve performance on it.
We have a benchmark here that actually performs WORSE than our hotdog bench.