There is a strange contradiction in the modern job market.
Remote work theoretically opened the country.
If a company in Boston needs a front-end engineer and I live in Florida, geography no longer has to end the conversation. A good job is no longer limited to the people willing to move within commuting distance of an office.
That sounds like a much larger opportunity.
It is.
It is also a much larger pile of résumés.
In February 2026, only about 9% of paid U.S. job postings on LinkedIn were remote, yet those jobs attracted 37% of applications. Overall U.S. hiring on LinkedIn was still 23% below its February 2020 level.
LinkedIn had already found something else interesting: compared with the pre-pandemic market, job seekers were submitting roughly twice as many applications even after the number of jobs relative to job seekers had returned closer to historical levels.
Greenhouse sees the same pressure from the employer side. Its 2026 benchmark work analyzed hundreds of millions of applications and found that applications per recruiter had increased dramatically while recruiting organizations became smaller. Its 2025 data showed applications per opening rising from an average of 28 in 2021 to 95 in 2025.
And averages are averages.
A desirable nationwide remote technology job can blow past them almost immediately.
So I have been asking a different question.
Not:
How do I apply to more jobs?
Not even:
How do I write a better résumé?
The question I care about is:
How do I make it to the top of the stack?
That wording is intentional.
Because somewhere between clicking Apply and talking to a hiring manager, I become an object in a stack.
And that stack is increasingly being processed by software before it is processed by a person.
Words have become cheap
Generative AI has done something remarkable for job seekers.
It can take a job description and produce a polished résumé.
It can write a cover letter.
It can identify keywords.
It can clean up grammar.
It can rewrite experience to sound more relevant.
It can do all of that in seconds.
Which means everybody else can do it too.
Greenhouse reported that 22% of surveyed candidates were already using bots to apply automatically, while 26% said AI had actually made it harder for them to stand out.
That is the paradox.
AI makes it easier to create a good application while simultaneously reducing the amount of information contained in the fact that an application looks good.
A beautifully written cover letter used to require some effort.
Now it requires a prompt.
A carefully tailored professional summary used to suggest that somebody had read the job description.
Now it may suggest that software read it.
This does not make AI bad.
I use AI constantly.
The mistake is believing that using the same tool everybody else has somehow creates differentiation.
It does not.
It raises the baseline.
When everybody can generate polished claims, claims lose value.
Evidence gains value.
There isn't really one stack
When I say I want to reach the top of the stack, I am increasingly convinced that there are several stacks sitting on top of each other.
The first is the machine stack.
Can an ATS understand who I am?
Does my résumé clearly contain the skills the employer is actually looking for?
Are Angular, TypeScript, accessibility, UX engineering, front-end architecture and the other relevant capabilities expressed plainly enough that software does not have to infer them?
Is the document structured sanely?
Is the terminology consistent?
Am I making the computer's job unnecessarily difficult because I wanted the résumé to look clever?
This part still matters.
I am not advocating abandoning the résumé or ignoring ATS optimization.
Quite the opposite.
The résumé has to survive.
But survival is not differentiation.
After that comes the attention stack.
A human being finally sees me.
Maybe for ten seconds.
Maybe five.
What happens?
Do they see another Senior Front-End Whatever Person with Angular, JavaScript, TypeScript, Git, Agile and excellent communication skills?
Or is there something that makes them stop?
That is a completely different UX problem.
Then comes what I think may be the most important layer now:
The trust stack.
Can the things I claim be verified?
The portfolio cannot just say I am good at building products
This realization has changed what I think a technical portfolio should be.
For years, portfolios were essentially galleries.
Here are some screenshots.
Here is a project.
Here is the technology I used.
Here is a paragraph explaining that I am passionate about solving difficult problems.
There is nothing inherently wrong with that.
But in a market drowning in claims, I don't think it is enough.
If I say I am a front-end architect, I should be able to show architecture.
If I say I understand accessibility, the portfolio itself should withstand accessibility scrutiny.
If I say I care about performance, I should measure performance.
If I say I work iteratively, there should be evidence of iteration.
If I say I can ship software, something should be shipping.
That thinking gradually changed djamespoer.com from a portfolio into something closer to a working laboratory.
It is built in Angular and TypeScript.
It has a real deployment pipeline.
It has accessibility tests.
It has performance measurements.
It has bugs.
It has experiments that work.
It also has experiments that absolutely suck and get pulled back out.
That last category may be more important than it sounds.
Product work is not the ability to make everything you touch brilliant on the first try.
It is the ability to recognize when something isn't good enough, learn from it and change direction.
A portfolio that only contains perfect hindsight hides that entire part of the job.
Make the claim inspectable
One of the systems we have been building is called Thingamajob.
Its premise is simple:
Before tailoring a résumé to a job, establish what is actually true.
Career evidence is gathered into a source-backed Library of Truth. A particular job description can then be evaluated against that evidence.
The point is not to have AI invent the best possible version of me.
The point is to let AI help identify the strongest truthful version of me for that particular opportunity.
That distinction matters more in an AI market, not less.
Another architectural decision separates my Work History from my Case Studies.
The Work History answers:
What did you do, where, and when?
The Case Study answers:
Why did you do it that way? What happened? What failed? What did you learn?
Those are different questions.
Stuffing both into one polished marketing page makes both harder to evaluate.
Buildometer takes the same philosophy further.
Instead of saying the portfolio is actively being developed, it attempts to measure the development.
What has shipped?
What remains?
What is the throughput?
What is being reworked?
The project is trying to turn its own development into evidence.
The product is documenting the process that creates the product.
That may be excessive for a portfolio.
Good.
That is partly the point.
Skills-based hiring makes this more interesting
There is a larger industry movement happening at the same time.
LinkedIn's 2025 recruiting research found that 93% of talent-acquisition professionals considered accurate skills assessment important to improving quality of hire. Companies doing the most skills-based searching were 12% more likely, by LinkedIn's measure, to make a quality hire.
That is a useful signal.
Employers increasingly say they want to know what candidates can actually do.
But a résumé remains mostly a list of assertions about what somebody can do.
That creates an opportunity.
Don't simply tell them.
Reduce the cost of verifying it.
If my résumé says Angular, the portfolio is Angular.
If it says accessibility engineering, use the site with a keyboard.
Inspect the semantics.
Turn on reduced motion.
Look at what happens on a phone.
If I say I understand design systems, architecture, AI-assisted development, performance or product iteration, the work should leave fingerprints.
A hiring manager should not have to believe my adjectives.
I want the nouns and verbs to do the work.
The best application may be the one that escapes the application pile
There is an uncomfortable piece of data here too.
Ashby analyzed more than 38 million applications across 93,000 jobs.
About 94% of applications were ordinary inbound applications.
Only about 1% came from referrals.
But those tiny categories behaved very differently once hiring started.
Approximately 3% of inbound applicants reached an interview.
About 40% of referred candidates did.
That does not mean everybody needs to start harassing strangers on LinkedIn asking for referrals.
It means there is another form of signal at work:
someone or something has already reduced uncertainty.
A referral says, at minimum, somebody is willing to attach a little of their own credibility to this person.
So perhaps the portfolio has another job.
It needs to be easy to forward.
If somebody I know encounters a position and thinks of me, I want one URL to give them.
Not:
"Trust me, James is really good."
Something closer to:
"Take a look at this. You'll understand why I thought of him."
That is a very different objective from building a portfolio whose primary goal is to win design awards.
Signal per second
I keep coming back to one phrase:
signal per second.
The recruiter has too many applications.
The hiring manager has too many responsibilities.
Neither one owes me twenty minutes of careful investigation.
So the design problem becomes:
How quickly can I communicate something meaningful?
And then:
How quickly can I prove it?
That is why visual differentiation still matters.
Not decoration for decoration's sake.
Attention is the entrance.
Proof is what has to be behind the door.
A memorable carousel, a strange visual system, Creative Town, a ridiculous Buildometer muscle car, animated layered artwork or a dog eventually running through the portfolio chasing biscuits may sound like exactly the wrong response to a serious job market.
Maybe.
But if a thousand technically competent applicants present themselves through nearly identical résumé templates, LinkedIn profiles and Bootstrap-looking portfolio grids, looking like everybody else becomes its own risk.
The challenge is not to be weird instead of credible.
The challenge is to be:
memorable enough to inspect, and credible enough to survive inspection.
Those are not opposites.
That intersection is where I want the portfolio to live.
AI should increase the need for authorship
There is another reason this matters.
As AI gets better at producing code, writing and design, employers are going to have a harder time determining what a candidate actually contributed.
I think authorship becomes more important, not less.
What decision did you make?
Why?
What constraints did you notice?
What did you reject?
What happened when reality disagreed with the model?
Where did your judgment enter the system?
AI can produce alternatives at extraordinary speed.
It still cannot be accountable for my taste.
It cannot decide that a slide we spent hours building is ugly and pull the damned thing down because I hate it.
That is still my job.
The ability to generate is becoming abundant.
Judgment remains scarce.
So how do I get to the top of the stack?
I don't think there is a trick.
If there were, everybody would use it and the trick would stop working.
I think the answer is a system.
Make the résumé machine-readable and truthful.
Tailor it aggressively to the job without tailoring reality.
Make the professional identity consistent enough that a search engine, ATS, recruiter and hiring manager encounter the same person.
Create something memorable enough to interrupt the scanning pattern.
Then put evidence immediately behind the claim.
Expose the work.
Expose the thinking.
Measure things.
Show receipts.
Make it easy for another human being to forward your work with confidence.
And keep building in public so the portfolio describes who you are now, not merely who you were at the last company that gave you a title.
Will that guarantee that I beat two thousand other applicants for a remote position?
Of course not.
That would be bullshit.
But sending application number 2,001 into the same pile using the same strategy as everyone else does not seem particularly compelling either.
So this is the experiment.
Instead of asking how I can produce more applications, I am trying to produce more signal.
Instead of telling employers that I can build products, I am building one they can inspect.
Instead of waiting for an interview to explain how I think, I am trying to leave the thinking visible.
And instead of treating the modern hiring stack as something I have to trick my way through, I am treating it like every other difficult UX problem:
Understand the system.
Understand the humans inside it.
Remove friction.
Create a reason to continue.
Then prove the promise was real.
Maybe that's how I get to the top of the stack.
We'll see.