What it does
It carries further than expected. The two places it breaks are reproduced in the benchmark: two-decimal probabilities make large result sets tie at 0.99, and a popular batching proxy flips passing rows into failing ones.
Benchmarks & research
Databases sort by numbers and Jev hands back probabilities, so sorting search results by them is tempting. This measures how far the temptation carries.
Only you can see your notes.
Screenshot unavailable. Open the experiment ↗
It carries further than expected. The two places it breaks are reproduced in the benchmark: two-decimal probabilities make large result sets tie at 0.99, and a popular batching proxy flips passing rows into failing ones.
Try this project's comparison method before using AI to sort your shop's search results. Ask a developer to compare the AI's order with choices made by people. Include difficult searches, not only easy ones. The maker found that good results on simple examples did not carry over to harder shopping searches.