Play the video in the post belowPost on X
- Views
- 25,439
- Likes
- 75
- Reposts
- 4
Read from X on . Counts change daily.
Kai rebuilt a job-hunting crawler around Jev after finding every company organizes its careers section differently. Starting from a company’s homepage, Jev identifies the entry point into the careers section, chooses which links on the page are worth following, recognizes when a page is an actual job posting, and scores each role it finds against the user’s own profile.
The same task with a general LLM handling each step took about 5 minutes in Kai’s earlier version. With Jev making the small decisions, it dropped to just over 20 seconds in testing. Kai frames the fit generally: automation that needs many small decisions along the way is where Jev’s faster responses and lower cost pay off, compared with calling a full chat model at every step.
Kai packaged the result as a skill, jev-job-hunter, and posted a demo video.



