Turning job-search uncertainty into an executable system
The problem is rarely a lack of listings. It is deciding which role is worth acting on, whether personal experience forms credible evidence, and what to do next.
Data: the July 17, 2026 full refresh and qualification queue. Tests rerun on July 21, 2026.
These are engineering-state metrics for a private, single-user system—not user counts, placement rates, or commercial conversion.
More information can make action slower
Most job aggregators leave the real work to the candidate: open every site, reread each JD, assess location and seniority, recall which experience proves the fit, then rebuild the resume, application note, and interview stories.
None of these decisions is especially difficult alone. Compressed together, they create a persistent feeling that something has been missed.
Hard gates before relevance
I defined the product as a local-first personal decision system, not a public job board. Policy signals and market heat may help discover directions, but they cannot override location, work authorization, minimum experience, seniority, or required skills.
The core product unit became a traceable action chain: full JD, qualification decision, personal evidence, risk, role-specific materials, and a next step.
Absorb information density in the back; deliver action in the front
The system must know when to say “no”
A role needs a full JD, location, and official entry point before qualification. Failed sources retain historical records.
Location, authorization, experience, seniority, and required skills take precedence over preferences and semantic similarity.
Direct experience, transferable ability, and unresolved risk stay separate. “Related to” cannot become “did.”
Recommendation rationale, resume, application note, and interview preparation share the same verified facts.
The system prepares materials but never handles CAPTCHAs, accepts privacy terms, or performs final submission.
Outcomes reorder qualified roles; they never restore a role blocked by a hard requirement.
Successes, failures, and boundaries remain visible
The latest full refresh produced 47 source executions: 43 succeeded and 4 failed. It aggregated 1,667 records and completed deduplication and qualification assessment for 1,666. All 10 currently actionable roles have reviewed materials.
Desktop and mobile critical paths were reverified. Failed sources were not hidden or replaced with new sources for the portfolio.
A public portfolio is not a public private-job-search system
This case study uses only public job fields, redacted verified metrics, and system structure. It does not read the working version that contains private job-search information.
Safe to show
- Public company, role, and JD fields
- Workflow, decision framework, and test methodology
- Redacted screenshots and verified aggregate metrics
Never public
- Name, email, phone, resume, or application materials
- Personal model, fit rationale, or real application status
- Private company judgments, secrets, or private builds
Validate ranking with real outcomes—not more sources
Next: repair and reverify the four failed sources. Once enough application and interview outcomes exist, evaluate whether the ranking actually improves action efficiency.