A career has edges. The schema only stores nodes.
I spent years in commercial kitchens and over a decade running a one-man construction business before I ever shipped software professionally. Here is what those years look like inside an applicant tracking system:
Sous Chef, Illahe Hills Country Club, 1996-1997
Owner, Alger Construction, 2006-2018
Two rows. Six fields. Everything that made those years worth putting on a resume in the first place lives in the space between them, and there is no field for the space between them.
That is not a bug in any particular vendor's product. It is what happens when a career gets stored.
Compression Is the Point, and Compression Has a Cost
A resume is already a lossy representation of a working life. An applicant tracking system compresses it again into titles, employers, dates, skills, and years of experience. Recruiter searches narrow it further. Screening criteria narrow it again. By the time a hiring manager sees a candidate, a great deal of what made that person worth meeting has been discarded by systems doing exactly what they were built to do.
Hiring at scale requires structure. Nobody sensible is arguing otherwise.
But compression is only lossless when the thing being compressed matches the schema. Software engineer becomes senior software engineer becomes staff software engineer. The titles line up. The keywords line up. The progression survives the trip through the pipeline nearly intact.
For a career that moved sideways, the same pipeline behaves very differently. It preserves the least interesting facts and drops the reason those facts belong on the same page.
The Resume Contains the Nodes. Something Has to Explain the Edges.
Running a construction business teaches you schedules, budgets, vendors, dependencies, unhappy customers, and what it costs to discover a problem after the wall is closed. Working a station on a busy line teaches you sequencing, preparation, and how to communicate when six things are failing at once and none of them are your fault. Managing a distributed technical team a decade later draws on both, whether or not the job description has a checkbox for it.
Systems fail. Priorities collide. Resources are finite. People misunderstand each other. Plans meet reality. Learning to operate under those conditions is transferable, and it is the actual content of a nonlinear career.
A person reading a resume can sometimes reconstruct those connections. A parser has a harder problem, because almost everything it has been given is categorical. It is very good at representing what fits in a field. It has nowhere to put "running a small business taught me something about operational risk that later made me better at managing engineers."
The information exists. The schema has no column for it.
This is why the quiet disappearance of the cover letter is more consequential than it looks. I understand why nobody misses them. Most were three paragraphs about being passionate about synergizing solutions in a fast-paced environment, and they earned their reputation. But the format was carrying a load that nothing replaced. The more nonlinear the candidate, the more of their value lives in relational information, and the application form keeps getting narrower precisely where that information needs to go.
We tell people to bring their whole selves to work while steadily reducing the bandwidth available to transmit one through the front door.
The Strongest Objection Is Not Efficiency
The honest counterargument to everything above is not that narrative is slow to read. It is that narrative is where bias lives.
Structured screening was partly a correction. Unstructured hiring, the kind that runs on interesting backgrounds and good conversations and a hiring manager's instinct about fit, has a well documented tendency to favor candidates who resemble the person doing the evaluating. "This person's unusual path is fascinating" has historically meant "this person reminds me of me." Fields, rubrics, and consistent criteria were a response to a real failure, not an accident of software convenience.
So the argument cannot be that we should bring back the cover letter and read it with an open heart. That is a request to reopen a door that was closed for a reason.
The argument is narrower, and I think it survives the objection: the correction discarded a category of information instead of evaluating it consistently. Those are different problems with different fixes. Consistency is achievable for prose. Ask every candidate the same question. Score the answers against the same rubric. Read them at the same stage. That is structured data that happens to arrive as sentences.
What we built instead was a process that solved the bias problem by deleting the input.
The Keyword Arms Race Is a Symptom of the Deletion
The predictable response has been optimization on both sides. Candidates tailor resumes to postings. Tools score keyword alignment. Generative models rewrite bullets in the vocabulary of the job description, while recruiters deploy increasingly sophisticated matching and ranking on the other end.
None of that is unreasonable in isolation. If a company says "developer enablement" and my resume says "developer education," making that equivalence explicit helps humans and machines alike.
Taken far enough, though, the process gets strange. One model rewrites a person's career into the vocabulary most likely to satisfy another model deciding whether a human being should ever see it. We have built a translation layer between two parties who could have had a conversation.
And the translation runs in one direction only. Optimizing for machine legibility strips out exactly the information companies claim to be looking for. Adaptability is messy. Cross-disciplinary experience is messy. Pivots are messy. That messiness is usually where the signal is.
A Hiring Process Is a Statement of What an Organization Preserves
It is easy to blame the ATS, because it is the visible machinery. It is also the least interesting part of the story.
Someone decides which fields matter. Someone writes the screening questions. Someone marks a qualification as required rather than preferred. Someone chooses how recruiters query the pool, and whether a hiring manager sees fifty candidates, ten, or only the five that most closely resemble the profile used to write the posting. And someone decides you will never be told which of those things happened to you.
The resulting process is not neutral. It is a record of what an organization decided was worth keeping.
A process that preserves titles, years, and exact skills while discarding explanation will reliably favor careers legible through titles, years, and exact skills. A process that rewards similarity to a predefined profile will filter out candidates whose value comes from unusual combinations, because unusual combinations are by definition dissimilar.
For plenty of roles that is fine. Nobody wants an airline hiring pilots because the career story was compelling.
It is harder to defend when the same organization says it is looking for adaptable leaders, systems thinkers, creative problem solvers, and people who challenge established assumptions. Those are all contextual qualities. Context is the first thing lossy compression throws away.
We Do Not Need Better Resume Optimizers
I have not solved hiring, and candidates still have to operate inside the system that exists. Make transferable skills explicit. Describe outcomes in the language of the role. Find a human being when you can. A nonlinear career does not become legible because we insist it deserves to be.
But another generation of tools that help candidates compress themselves more efficiently is not a fix. It is better compliance with the thing causing the problem.
The smaller fix is available today and costs almost nothing. One field, asked of every applicant, scored the same way: what connects these experiences, and what did you carry from one to the next? Two hundred words. Read before the rejection, not after the offer. It is consistent, it is auditable, and it gives relational information somewhere to exist inside a structured process rather than outside it.
If an organization genuinely believes that unconventional backgrounds and adaptable thinkers create value, its hiring process needs a place to put them. Not in the culture page. In the schema, where decisions actually get made.
Otherwise, "bring your whole self to work" carries an undocumented prerequisite:
First, make sure your whole self fits in the fields.
P.S. I have also prepared my curriculum vitae in Latin. Thirty-five years of experience, zero recoverable keywords, and one very specific classics-major-turned-recruiter out there who will open it and say: hic est candidatus quem exspectabam.
Top comments (2)
I actually wrote an app for it. My finding has been in the modern ecosystem that automation removed all the variables, that made a cv stand out. Recruiters literally just read a 1 paragraph summary of you, spat out by AI... So if they're going to read a summary, have it be relevant. I made Resume-Helper as an open-source tool, so you write your own info, it stores to a db as blocks of text, covering all the various skills, talents and experiences you've had, then compiles a resume per-job posting and red-teams it, to see if you meet all requirements and what sets you apart, whether it's relevant, etc. So you dont manually craft the 100 resume's that never get a callback and the rating system gives you a pretty solid estimate on how relevant your CV is to the job posting from a third-person perspective.
I dont wanna say it works, but I did get my first interview in over a year after just 10 resumes sent out, though that could just be luck. That being said, out of all the companies I applied to, it is the one my skills are most relevant for... So there's that, atleast it translated well.
dev.to/unitbuilds_cc/resume-advice...
If you wanna read about it.
github.com/UnitBuilds-CC/Resume-He...
Or want to give it a try? Just add it's MCP to your LLM and you're good to go!