Portfolios Beat Degrees: 4,192 Jobs Data

Open any job board and you will find a thousand opinions about the future of work. We had a different idea: skip the opinions and read every job listing on Loker Dollar end to end.

All 4,192 of them. Active and inactive. Across every category, every remote type, every pay range, as of mid-May 2026.

What follows is what employers are actually writing, not what vendors and pundits keep saying they are. The gaps are large. They are uncomfortable. And they are pointed at almost everyone reading this.

This pillar covers seven findings in depth. Four audience-specific companion posts dig deeper for each reader profile — employers, B2B sellers into hiring teams, tech job seekers, and non-tech job seekers — and are linked inline at the end of each finding.

Finding 1: Portfolios Beat Degrees 3.3 to 1

TL;DR. Across 4,192 listings, 591 (14 percent) ask for a portfolio, GitHub, or case study. Only 179 (4.3 percent) require a degree. The credential is collapsing as a filter and almost nobody is announcing it.

Three numbers anchor this finding:

• 591 listings explicitly request a portfolio, GitHub profile, or case-study attachment.
• 179 listings require a degree (bachelor, master, PhD, or named field of study).
• 612 listings — almost 15 percent of the board — use proof-of-work language: "show your work," "examples of past projects," "previous shipping experience," "we want to see what you have built."

The ratio that matters is portfolio-to-degree: 3.3 to 1. For every employer asking for paper, more than three are asking for evidence.

Equally telling: only four listings out of 4,192 explicitly say "no degree required." The market has not switched from "degree required" to "degree explicitly not required." It has switched to a quieter middle position — the credential is no longer named at all. Hiring managers learned that asking for proof of work tells them more than asking for a degree ever did.

What this means in practice depends on which side of the table you sit on.

If you are a job seeker, the math is brutal and clean. A certificate without a public artifact does not survive contact with this board. A modest portfolio with one shipped piece beats a long resume with no demonstrable work. Stop spending the credential premium until you have proof — the buyer has moved.

If you are a hiring manager, your interview rubric is the test. If it grades degree first, scoreable artifacts second, you are running a 2015 process against a 2026 candidate pool. The peer companies whose listings you are competing against have already moved past the credential as a filter.

Caveat to register: this is a Loker Dollar snapshot. Boards that index more heavily into regulated industries — medicine, law, finance compliance — will show different ratios. Our data tilts toward tech, design, content, and remote operations work. Translate the directional signal carefully into your sector.

→ Deeper read for employers: The 4,192-Job Employer Playbook. For tech seekers: Tech Job Seeker Strategy.

Finding 2: "Replace" Outpaces "Augment" 1.6 to 1

TL;DR. Vendors say "augment." Real job descriptions say "replace" — 1.6 times more often. Whether or not the headcount reductions ship, the candidate pool is reading the words you actually wrote.

For two years, vendor decks have insisted that AI will "augment" workers, not replace them. The job market is not writing that way.

Across the body of all 4,192 listings on the board, "replace" appears in 37 descriptions while "augment" appears in 23 — a 1.6 to 1 ratio. The numbers are small in absolute terms, but the direction is consistent: when employers reach for a verb to describe what AI is doing inside their workflows, they reach for the harder one more often than the softer one.

That framing leaks. It leaks into how interview loops grade for "AI-readiness." It leaks into compensation bands that quietly shrink for roles whose tasks are now partly automated. It leaks most of all into the body of the job description itself, where the cost of being honest is zero and the cost of being soft is being misread by candidates who have learned to look for the harder word.

Three implications:

• If you sell yourself as a worker AI will "augment" while your would-be employer is writing "replace," your interview is going to feel different than you expected. Re-read the listing for verbs, not job titles.
• If you write listings on behalf of your employer brand, audit the verbs your peers use in the same role. "Replace" without a follow-through ships layoffs you did not plan. "Augment" while your competitors say "replace" ships an interview pipeline that is mispriced against the market.
• The companies most likely to write "replace" are not always the most AI-mature. They are often the most cost-pressured. Whether or not the AI works, the cost narrative is already running. Plan for it.

Caveat: 60 listings out of 4,192 mention either verb. The vast majority say neither, because most descriptions do not editorialize about AI's effect on labor. The 1.6 ratio is a directional signal from the listings that do take a position — not a population estimate.

→ Deeper read for employers: Employer Playbook. For non-tech seekers: Non-Tech Survival Guide.

Finding 3: AI as Baseline Skill Is 2.6× Bigger Than AI as Specialty

TL;DR. Only 258 listings carry an AI title. But 673 non-AI titles name AI somewhere in the body. The market treats AI as baseline, not specialty — 2.6 to 1.

The popular story is that "AI jobs" are exploding. The data tells a quieter, more interesting story.

258 listings out of 4,192 (6.2 percent) carry an AI-related title — AI Engineer, ML Engineer, Prompt Engineer, GenAI Lead, NLP Specialist, and a handful of variants. That number is real and growing. It is also small compared to where AI fluency is actually being demanded.

673 listings whose titles have nothing to do with AI — Software Engineer, Product Designer, Growth Marketer, Operations Manager, Content Lead, Customer Success Manager — name AI somewhere in the body. They write things like "comfortable using AI tools in your day-to-day," "familiar with prompt engineering," "able to leverage Copilot and ChatGPT for productivity," or "we expect every hire to bring AI fluency to their craft."

The ratio is the headline: AI as baseline expectation outnumbers AI as job title 2.6 to 1. Tools-by-name (Cursor, Copilot, Claude, ChatGPT, v0) appear 133 times across the board. "Prompt" appears 109 times. None of those words are job titles. They are baseline skills attached to ordinary roles.

Three follow-on observations:

• The market has decided AI is closer to "version control" than to "kubernetes." Not every role builds it. Every role is increasingly expected to operate inside it.
• Specialist AI roles will keep growing in absolute terms. They will also keep shrinking as a share of all AI-relevant roles, because every non-AI role on the board is quietly absorbing AI into its day-to-day.
• The candidates being filtered out hardest right now are not the ones with no AI title. They are the ones with no AI mention anywhere on their CV, in a market where 22 percent of all listings expect at least baseline fluency in something AI-adjacent.

The practical move: if your CV says nothing about which AI tools you use in your actual work, you are losing to candidates whose CV does — in roles that have nothing to do with AI on paper.

→ Deeper read: Tech Job Seeker Strategy and Non-Tech Survival Guide.

Finding 4: The AI Title Pay Inversion

TL;DR. In our sample, AI Engineer / ML Engineer roles average $54,400 pay min. AI Trainer roles average $75,231. Senior+ roles average $124,462. The "AI" label, on its own, pays less than "senior" in any field.

This is the data point that should reshape how anyone thinks about chasing an "AI" title.

Of 1,488 listings that publish a stated salary range (35 percent of the board), 35 are AI Engineer or ML Engineer titles. Their average minimum yearly pay in USD is $54,400. Another 33 are AI Trainer, annotator, RLHF reviewer, or AI evaluator roles. Their average minimum is $75,231 — nearly 40 percent higher than the AI Engineer average. And 23 of those 33 AI Trainer-style roles (70 percent) require no coding at all.

The 840 listings carrying any senior, staff, principal, or lead designation average $124,462 — more than twice the AI Engineer average. Seniority in any field still pays better than the AI label.

Three reasons the inversion is plausible (not just a sampling fluke):

• "AI Engineer" is a young title with a wide salary band. Many listings using it are mid-level or early-senior, while the most expensive ML talent is increasingly hiding under "Research Scientist," "ML Platform Lead," or no AI-flavored title at all.
• "AI Trainer" sounds entry-level but in practice requires deep domain judgment — a lawyer reviewing legal answers, a doctor reviewing clinical answers, a senior writer evaluating tone. Frontier labs are paying for hard-to-replace human expertise, not for the title.
• The compounding effect: AI Trainer roles are often contract-and-remote, with hourly rates that annualize generously. AI Engineer roles in this dataset skew toward full-time at SMB compensation bands.

Caveat to register loudly: sample sizes are small (n=35, n=33). A different snapshot could move the average by 15-20 percent. The inversion is a directional signal worth investigating, not a binding benchmark.

The practical move: do not optimize a career for the AI title. Optimize for compounding domain depth plus public proof of work plus the seniority track. The market is paying for the combination, not the label.

→ Deeper read: Tech Job Seeker Strategy and Non-Tech Survival Guide.

Finding 5: Senior-to-Junior Is 4.4 to 1

TL;DR. 840 listings target senior/staff/principal/lead. 193 target junior/entry/intern/associate. The talent shortage you keep hearing about is one your hiring pipeline shaped.

Of 4,192 listings on the board, 840 carry an explicit senior, staff, principal, or lead designation. 193 carry a junior, entry-level, intern, or associate marker. That is a 4.4 to 1 ratio in favor of experienced hires.

Two interpretations matter:

• The market did not stop needing juniors. It stopped advertising for them. Many junior hires today happen through internal apprenticeships, bootcamp partnerships, or referral pipelines that never touch a public listing. The board reflects the public-facing slice, which is now overwhelmingly senior.
• The same companies that publicly lament a "talent shortage" stopped buying entry-level talent at the public-listing layer. Pipelines do not refill themselves. The senior premium gets larger every year you skip a junior cohort.

The candidate-side reality is brutal: anyone trying to break in is competing against a pipeline four times wider at every level above them, and the few junior listings that do appear close fastest. Inactive-job analysis shows junior roles close out within days, not weeks. Speed of application matters more for juniors than for seniors.

The employer-side reality is structural: every quarter you do not refill the junior intake is a quarter you have postponed, not avoided, the cost of replacing senior departures. The cohort math always wins eventually.

→ Deeper read for employers: Employer Playbook. For tech seekers: Tech Job Seeker Strategy.

Finding 6: The AI Contract Migration Has Already Started

TL;DR. Across the board, 30 percent of listings are contract or freelance. Inside the AI category, that share rises to 41 percent. Your best AI-capable employees are running the same math you are.

Across all 4,192 listings, 1,271 (30 percent) are contract or freelance, and 2,856 (68 percent) are full-time. Inside AI-relevant listings, the split tips toward contract: 132 of 321 AI-tagged roles (41 percent) are contract or freelance.

The compensation arithmetic helps explain it. Full-time pay min averages $96,465 across listings that disclose pay. Contract pay min averages $84,446 — the headline reads "contract pays less." But the headline is misleading. Contract pay is for one engagement, while a full-timer's number absorbs every adjacent cost: payroll tax, healthcare, benefits, ramp-up time, internal politics. Once a contractor's pipeline holds three concurrent clients, the math tips fast.

For AI-capable employees, the leverage is sharper still: AI work has a high per-engagement value ceiling (the unit of output is often discrete and shippable), tools-of-the-trade are portable, and the buyer base is large and global. The contractor pipeline assembles faster for an AI-capable mid-careerist than for almost any other profile in tech.

Three downstream effects worth watching:

• Companies that retain AI talent will need to compete on something other than base pay: equity, mission, problem quality, or operating cadence.
• Companies that fail to retain AI talent will pay 1.5-2x more on the contract side to get the same hours back — and the contractor controls scheduling.
• The contractor migration creates a flywheel for the AI tooling market itself, because contractors are price-elastic buyers of productivity tools in ways salaried employees are not.

The practical move for employers: if your AI-capable people are eligible for $84K-plus on contract right now, base pay is no longer the conversation. Pipeline quality, autonomy, and equity are. For AI-capable employees: the option value of going contract is already inside your compensation package, whether or not you exercise it.

→ Deeper read for employers: Employer Playbook. For B2B sellers into hiring teams: B2B Seller Data Pack.

Finding 7: What the Listings Quietly Say

Six smaller signals that compound when read together:

• Communication vs creativity, 5.6 to 1. "Communication" appears in 625 listings as a required skill. "Creativity" in 111. The most-named soft skill is also the cheapest to grade — and the rarest interview actually tests for it well.
• Specialist vs generalist, 11 to 1. 281 listings call themselves "specialist" roles in the body; 25 call themselves "generalist." Lead with the niche. The generalist range reveals itself in the second meeting.
• Domain expertise asked openly. 77 listings explicitly want SME or domain experience. 609 name a specific sector — journalism, finance, supply chain, healthcare, legal, accounting. The market is paying for "engineer who understands logistics," not "engineer who once worked near logistics."
• Pay opacity at 65 percent. Only 1,488 of 4,192 listings publish a salary range. The remaining 65 percent are pay-opaque. Whatever your candidate retention story is, your candidates are already pricing the silence.
• Worldwide jobs stay open 1.8x longer than regional. Worldwide remote roles have an 82 percent active rate (1,554 of 1,894). Regional remote roles run at 45 percent (1,037 of 2,298). Global gigs persist; local gigs close fast. Application speed matters more in the regional band.
• 97 percent of the board is under 90 days old. Any candidate or seller strategy that takes more than a quarter to execute is chasing a moving target. The market refreshes itself almost completely every quarter.

→ Deeper reads: Employer Playbook, B2B Seller Data Pack, Tech Job Seeker Strategy, Non-Tech Survival Guide.

Methodology Note

Every number in this piece is computed against the live Loker Dollar production database (bf92ac7b-b18d-4416-85fc-08764b3fe8c1), against 4,192 job listings — every category, both active and recently closed — as of mid-May 2026. Pay averages are computed only over the 1,488 listings (35 percent) that publish a stated salary range; the remaining 65 percent are pay-opaque and excluded from pay math. Where sample sizes are small (AI Engineer n=35, AI Trainer n=33, augment/replace verbs n=23/n=37), the finding is reported as a directional signal rather than a benchmark.

The full live dataset, refreshed weekly, is searchable for free on Loker Dollar. Every job in this analysis sits behind the same UI you would use to apply.

FAQ

How was the 4,192-job dataset assembled?

Every active or recently closed listing on the Loker Dollar production database as of 2026-05-18. 2,591 are active, 1,601 are closed. 97 percent were posted within the last 90 days. No paid sample, no synthetic data, no scraped third-party board — only listings ingested through our own pipeline.

Why does AI Trainer pay more than AI Engineer in your sample?

Two structural reasons and one statistical one. Structural: AI Trainer roles at frontier labs require deep domain judgment from senior practitioners (law, medicine, writing) and are increasingly priced as expert-witness work, not entry labor. AI Engineer titles in our snapshot skew toward SMB compensation bands. Statistical: with n=35 and n=33, a single high-paying outlier can shift the mean by several percent. Treat the inversion as a directional signal, not a binding benchmark.

Is "portfolio beats degree 3.3 to 1" true across all sectors?

No. The 3.3 ratio reflects this board's mix, which tilts toward tech, design, content, and remote operations work. Regulated sectors (medicine, law, finance compliance) still index harder on credentials and will show a different ratio on a different board. The directional point — that proof of work is rising as a filter relative to credentials — holds broadly in the categories Loker Dollar covers most densely.

Why exclude 65 percent of listings from pay math?

Because they do not publish a pay range. Imputing pay from job title alone would invent numbers the market did not commit to. We chose visible signal over fabricated coverage. The 35 percent that do disclose are the basis for every average in this piece, and we say so each time.

Are you saying degrees are worthless?

No. We are saying the public filter has shifted. A degree paired with a shipped portfolio outperforms either signal alone in this dataset. A degree without proof of work loses to a portfolio without a degree at a 3.3 to 1 ratio when both candidates apply to the same listing. The market is not anti-credential. It is pro-evidence.

What is the single highest-leverage move for a job seeker after reading this?

Ship one piece of public proof of work in the next 30 days, using at least one of the AI tools your industry names by hand. Portfolio is the new filter. The tool-by-name mentions (Cursor, Copilot, Claude, ChatGPT, v0) appear in 133 listings — make sure your artifact shows one of them in real use, not just listed as a "familiar with" line on a CV.