The problem

A posting says "remote", but can you actually be hired from where you live? Country rules, employment type, time zones and work authorization are buried in the text. Most people find out by applying and waiting.

JossJobs isn't a job board, an auto-apply bot or a resume blaster. It tells you which jobs are worth your time, and why.

What makes it different

Every eligibility or fit claim has to trace back to a quoted sentence in the posting. When the posting doesn't say, JossJobs says "uncertain" out loud rather than guessing. And an uncertain job can never outrank an eligible one.

The language model does one job, pulling facts out of the posting. The yes/no eligibility decisions are made by plain, auditable rules, which are cheaper, faster and easier to trust.

Extraction never sees who is asking, so each posting is read once and the result is reused for every user. That is the biggest lever on what the product costs to run.

Where it is now

A working web app and command-line tool, being tested by its builder ahead of a public launch. It reads company job boards on Greenhouse, Ashby and Lever, compares pay across currencies, and is covered by over a thousand backend tests. Its design choices are recorded in 47 written decision records.

How a posting becomes a recommendation

DiscoverPull new postings straight from company job boards, then normalize and remove duplicates.
ExtractClaude pulls the facts out of each posting. The same posting always yields the same facts, whoever is looking.
VerifyEach fact is checked against the posting text. Anything without evidence is marked unverified.
Assess eligibilityRules, not a model, check location, employment type, relocation and work authorization: eligible, uncertain or ineligible.
Score fit and rankFit against your CV comes last, shown as Strong, Good, Moderate or Weak. Eligibility always ranks first.