INSIDE JOBRIC

By Ric @ Jobric · September 2026

A match score with no reasoning attached is just an assertion. You're told a job is an 88% fit, you have no idea what that measured, and you're being asked to trust a number that refuses to explain itself.

August was the month we went at that directly. The score got sharper and the match card started handing you the evidence behind it.

A Jobric match arrives with a number and an explanation. The number is the fit score. The explanation is the AI fit analysis, the part that tells you why this role and not the forty others we looked at.

The score stacks several inputs together: how well the role lines up with your background, how it sits against the preferences you set, how much your skills overlap with what the posting asks for, and whether the seniority is right for where you are.

Two of those inputs got significant work in August.

Skill overlap now contributes as designed. It's the largest scoring factor after the core match and your preferences, and we identified and resolved an issue that was limiting how much of your actual skill set reached that comparison.

Seniority is now part of the calculation. That means real credit for a role that's a step up, a guard against putting an internship in front of a senior candidate, and a "stop showing me junior roles" preference that carries actual weight. Measured against production before release, 39 of 40 candidates gained matches.

Both ship switched off while we watch them against live results before turning them on. Changing two large scoring inputs simultaneously and hoping is not a plan, it's a wish.

There's a third piece queued alongside them. Scores have been clustering near the top of the scale, which compresses the distance between "strong" and "exceptional" into a narrow band and makes two genuinely different roles look alike to you. A recalibration that spreads scores across the full range, with the tier boundaries moved to match, is built and ready. It ships switched off for the same reason.

The match card shows its work.

Opening a match used to give you a number and a wall of prose, with no employer identity to anchor it and an explanation written partly in applicant-tracking dialect. Phrases like "strong semantic alignment with job description" tell you which system produced the answer and precisely nothing about what it means.

Now every match leads with the employer, their logo where we have it and a clean branded monogram of their initials where we don't. The analysis produces short, checkable statements, things like "Azure architecture, Principal Cloud Architect, 7+ years hands on," which render as a checklist inside the match and as up to three chips on the card face. The opened match splits into two zones, "Your fit" and "The role," with the strongest line of the analysis sitting beside the score as an opening verdict.

And we deliberately show you the one thing that doesn't fit, right next to the things that do, in neutral gray rather than a warning color. A gap is a fact, not an alarm. It's also the thing that makes the other three believable, because a match that claims you're perfect for everything is a match that isn't reading anything.

Matches analyzed before this shipped, which is most of what you're looking at, fall back into the same shape rather than reverting to the old layout.

A substantially larger job pool.

We rebuilt how we pull postings from three major applicant tracking platforms, which between them cover most of the large employers we track.

One spans 44 company boards and roughly 7,400 fresh postings a day, and its daily run now saves in batches as it goes and stops on its own terms before running out of clock. Another spans 36 boards, and we taught the intake to read how those boards express posting dates, which is text like "Posted 4 Days Ago" rather than an actual date. On the third, a fetch across all six live boards now finishes in 62 seconds, including the three largest, one of which carries 4,740 listings.

We sharpened the reporting behind all of it too, so "no new jobs today" and "this board needs attention" are now distinguishable at a glance. Six more boards surfaced that way and got repaired or retired, including one large employer that now contributes 1,834 postings on its own.

The intake pipeline also got faster: per-job overhead dropped from 213 milliseconds to 15, which takes a batch of 500 jobs from about 106 seconds to about 8. Faster intake means today's postings reach you as today's matches.

Wide Lens.

Your matches are scoped to the one career category we sort you into when we read your resume. That works until it doesn't. Someone with a technology background who also teaches a class gets sorted as a teacher and sees teaching roles, and a single categorization decision quietly shapes everything they ever see.

Wide Lens is a switch on the paid tier that widens your daily search across every fresh job we have. Out of category roles get held to a higher bar, not a lower one, specifically because we can't lean on category alignment to vouch for them. This is a wider net, not a looser one. Results arrive at two extra matches a day rather than ten at once, and each carries a Wide Lens label on every screen, with its score presented as its own kind of result rather than a weak version of the normal one.

And one for the privacy column: you can now delete your account yourself. There's a delete flow in Settings, three confirmations deep, permanent, with no grace period and no "we'll miss you" pitch in the confirmation email. Everything identifying you goes: name, email, sign-in identity, the resume file and its text, work history, certifications, location, coaching answers, guide messages. What remains is the anonymous record of which jobs were matched and how they scored, carrying no name, no email and no resume.

The monthly drawer of small things that were doing something faintly absurd.

Everything was sparkling. Match cards carried a premium shimmer effect, which sounds great until you learn it was running on essentially every card. When everything shimmers, the shimmer is just weather. It's now reserved for Exceptional Fit. The label underneath also used to read "AI Match," which told you which system generated the number and nothing at all about whether the job was any good. It now reads Good Fit, Strong Fit or Exceptional Fit.

Your RN license and your Excel proficiency were roommates. Credentials used to dissolve into the general skills list, sitting there companionably alongside whatever software you happened to list. They're now their own field, with the issuing body and whether it's state or board licensure. Nothing gets inferred, either: a Nurse Manager isn't treated as holding an RN license unless your resume says so.

Your work history was being summarized rather than read. Parsing now extracts each position with its employer and its start and end dates, including periods where you held two titles at once. Years of experience is calculated from those dates instead of totalled up by the AI, which means the number is checkable rather than asserted. Where your resume doesn't state a date, we leave it blank instead of confidently inventing one.

Search Health was blaming your filters for a crime committed three steps later. The explanation for a thin search used to stop at your preference filters, which put the whole shortfall on your preferences even when most of it happened downstream at the quality cut. It now walks the full path: roles looked at, roles ruled out, roles that cleared preferences, roles that scored well enough to show. Pointing you at a lever that wouldn't have helped is worse than saying nothing.

Switching the scoring work on. Skills, seniority and recalibration are built and waiting. Expected to go live once we've watched each against real matches, deliberately and one at a time.

Credentials counting for more than titles. A reference table of around seventy credentials, recording what each one actually proves, is built and testable but switched off. Nobody holds a professional license by accident, which makes it far stronger evidence of what you do than a string of titles your employers happened to choose.

A resume report. Groundwork so far. What we could and couldn't read off your resume is now recorded when we parse it, market data covers every career category somebody is actually in rather than only the popular ones, and your top market gaps are computed alongside. None of it is visible yet. It's being collected now so the report has real history to show the day it launches.

Look back at that list and most of it is one idea in different clothes.

A score that shows its evidence. A gap displayed beside the strengths instead of tucked behind them. A thin search that names which stage actually thinned it. A Wide Lens result labeled as a different kind of result rather than quietly mixed in.

That's not a feature, it's a posture. The product should be able to tell you what it did and why, because a recommendation you can't interrogate is just a company telling you to trust it.

Your job search should be able to explain itself. So should the thing running it.

The job market works for you now.

That's the update. Now go do something that isn't job searching.

Ric @ Jobric