jevIMAGE
Running Jev on 5 real datasets
The author used prompt_jev() in MotherDuck on five kinds of real data: agent logs, HN comments, sales calls, job postings, and regex parsing, with code and datasets included.
Mehdi Ouazza
@mehd_ioArchived 26 days agoOriginally posted 2026.10.06
Mehdi Ouazza @mehd_ioOriginally posted 2026.10.06
I spent a lot of time with Jev last week. Since prompt_jev() shipped in MotherDuck, testing it on that untapped text column is super easy. But what should you use it for? I took 5 real datasets to find out: 1. Agent log tracing: which tool calls were wasted? 2. Hacker News: what does HN think of each AI lab? 3. Sales calls: which ones match the use cases you serve? 4. Job postings: which cloud does each market run on? 5. Regex parsers: one question instead of 22 hand-tuned patterns Deep dive, code and datasets: https://motherduck.com/blog/jev-analytics-use-cases/
x.com/mehd_io/status/2107495137257332824Archived 2026.10.07 09:59