Job filters: what already works, what is still missing¶
Technical version: WING-196 requirement gap matrix
Status in one sentence¶
A solid foundation for job queries is tested (keywords, work model, fit level, source, ATS kind, dates, four job buckets, facets, sorting, paging, dedupe, redacted results). The full WING-196 epic is not finished.
Gaps that still need real data + UI¶
The technical matrix tracks each requirement as: model evidence needed, query work, UI work, tests, and owner lane. In plain language, still missing or partial includes:
- Career tier, confidence, and user corrections
- Discipline taxonomy
- Skill intent (own / grow / want / avoid) with proof strength
- Remote geography eligibility (not just a remote checkbox)
- Place, commute, and multi-location reality
- Original vs repost history
- Salary, bonus, and equity with honest assumptions
- Richer ATS support levels
- Company type, industry, cohorts
- Company health and layoffs with freshness
- Benefits / ERGs / insurance with provenance
- Federal path filters without inventing eligibility
- Application readiness
- Shared filter rail across apps with disabled states when a surface cannot support a control
- Total tax burden in comparisons (financially sensitive — must stay deterministic and sourced)
Rules that keep the work honest¶
- Unknown stays unknown. Missing evidence is never rewritten as “no” or “bad.”
- Model/query, location, and adapter lanes do not steal each other’s files.
- The existing frontend filter rail is the only UI integration target once backends land.
- Financial and tax rules need independent review, not a pretty guess.