job.army is the longest arc in this blog, a parked domain that became a living federal jobs board and then an engine that builds boards for any niche. It is also the project with the most honest unfinished business. This finale is the full accounting, the machine, the lessons, and the wall.
What actually got built
A plugin grown to thousands of lines around a jab_job post type with department, type, and location taxonomies, fed by a discovery engine of adapters, USAJOBS through its header-authenticated API with Fields=Full announcements, Adzuna, Jooble, generic JSON, and reluctant scraping, all emitting one standard shape into one import pipeline with strength-ordered deduplication. Around the content, an AI rewrite engine under strict fact-preservation constraints, JobPosting schema built from real meta with locations split properly and two fields honestly blank, a custom sitemap index, an expiry checker that asks before deleting, and a formatting pipeline that learned to read rendered output rather than trust pattern counts. Around the product, settings for everything, an AdSense readiness pack, backup and restore, documentation answering buyers’ questions in advance, and a companion theme, with job.army as the showroom.
The lessons with the longest reach
Source choice is destiny, the official API made this project possible and its absence would have made it the silver-site scraper wars at ten times the scale. Absence is ambiguous, the settings-wipe bug, whose fix, per-form field manifests, I now reach for in any partial-update design. Removals need checklists, the deleted-function fatal turned grep-before-and-after into reflex. Honesty has a floor even in metadata, blank schema fields over invented zips. And the hardest one, mitigation is not transformation, the AI rewriter reduced duplication but only original articles gave the domain an identity, aggregators earn their place by what they add, not what they restate.
A few things people ask me about this
What is the honest state of the AdSense wall? Rejected more than once, structurally, an aggregator reads as republished content regardless of formatting polish. The working strategy is original guides building traffic first, then reapplication, patience plus additive content, not cosmetic resubmission.
Was the universal product worth building before the first site succeeded fully? The engine work, yes, it is the durable asset either way. The sequencing lesson stands though, my own pattern of starting new builds before earlier ones reach profit is real, and this project is both its best output and its best evidence.
One thing you would tell someone starting an aggregator? Pick the niche by the quality of its best data source, and budget as much effort for what you will add above the feed as for the feed itself.
Where this leads
The board began because parking a domain earned pennies, and that thread deserves its own honest series, what domain investing actually is, what parking really pays, and why building on a name beats waiting to sell it. Those lessons are next.