Perspectives

The Entry-Level Resume Is Broken. Evidence-Based Hiring Is the Fix.

The Entry-Level Resume Is Broken. Evidence-Based Hiring Is the Fix. — Why the Skills Passport replaces the resume in a market broken on both ends.

The collapse of entry-level tech hiring isn’t only a talent problem — it’s a screening problem. Every recruiter knows the symptom: an open junior role posted Monday morning has hundreds of applicants by Friday, almost all of them holding similar degrees, similar GPAs, similar summer internships. The traditional resume can’t distinguish them, and AI-augmented applicant tooling has helped multiply the volume without improving the signal.

The same data that has new grads worried should also worry hiring teams. The pipeline is broken on both ends.

The Data Recruiters Should Be Looking At

Big Tech entry-level hiring fell 25% from 2023 to 2024, per SignalFire’s 2025 State of Talent Report. New graduates now account for just 7% of hires at the 15 largest tech firms, down from about 14% in 2019. Startups have moved in the same direction: new grads are under 6% of their hires, and new-grad hiring is down more than 30% from 2019.

Mid-level hiring grew in parallel. SignalFire found Big Tech increased hiring 27% for professionals with two to five years of experience in the same period. Startups hired 14% more in that range. Companies aren’t shrinking headcount across the board — they’re shifting it.

The Federal Reserve Bank of New York shows unemployment among recent college graduates up 30% between September 2022 and September 2026, versus 16% for all workers.

The headline reading is that AI is eating entry-level jobs. The truer reading is: companies have stopped paying the training cost that justifies entry-level hiring, because they can’t tell who’s worth training.

Why the Resume Has Stopped Working

A 22-year-old applicant in 2026 has, on paper:

  • A CS degree from one of nearly 4,000 accredited US institutions
  • A 3.4–3.9 GPA (compressed range, hard to discriminate)
  • One or two summer internships, often with self-reported scopes
  • A handful of class projects, indistinguishable from the next applicant’s
  • A LinkedIn profile shaped by AI-rewriting tools
  • A resume tailored by ChatGPT to the job description

The recruiter has, on their end:

  • Hundreds of applications for one role
  • An ATS that filters on rigid, exact-match criteria
  • About 7 seconds per resume on initial review (the eye-tracking studies have been remarkably stable: about 6 seconds in 2012, 7.4 in 2018)
  • No reliable way to verify any single claim

This is not a screening process. It is a lottery with a small skill-weighted component.

Worse: the rise of generative AI in candidate tooling means the noise floor has gone up. Every resume now reads professionally. Every cover letter is grammatically correct. Every LinkedIn summary is keyword-optimized. The traditional signal — that a candidate communicates clearly — has collapsed because AI handles that for everyone.

What Evidence-Based Hiring Actually Means

The replacement isn’t more interviews. Hiring teams have already pushed interview load toward the point of diminishing returns — they ran 42% more interviews per hire in 2024 than in 2021, according to Gem’s recruiting benchmarks, while Google’s own analysis found that four interviews were enough to predict with 86% confidence whether to hire someone. Adding screening interviews increases cost and slows time-to-hire without meaningfully improving prediction.

The replacement is front-loaded, calibrated evidence.

In practice, this means a few things:

1. Standardized assessment, not vendor-locked scores. A HackerRank score, a CodeSignal score, and an iMocha score are not directly comparable. A candidate scoring 720 on one platform may score 580 on another for reasons that have nothing to do with skill. Evidence-based hiring requires a normalization layer — a Skills Passport that aggregates across vendors and surfaces a single, calibrated composite.

2. Multi-pillar evaluation, not just code tests. Junior developer success draws on four pillars: domain (language proficiency, problem decomposition), AI fluency (prompt-to-spec, AI output evaluation), cognitive (reasoning, pattern recognition), and communication (written clarity, structured argument). A candidate strong in three pillars and weak in one is a different hiring decision than a candidate weak in three and strong in one — but a single LeetCode score collapses that distinction.

3. Recency-weighted scoring. A 2019 CodeSignal score isn’t worth what a 2024 score is — both because the candidate’s skills may have shifted and because the assessment instrument may have been recalibrated. Decay-aware composites reflect this.

4. Proof of work alongside test scores. A test result tells you what someone can do under timed conditions. A GitHub repository, a deployed project, or a public technical write-up tells you what someone does do without supervision. The combination tells you more than either alone.

5. Candidate-owned credentials. The credential should travel with the candidate, not stay locked in the vendor account or buried in an old ATS. This both serves the candidate (one passport, many applications) and serves the recruiter (verifiable evidence rather than self-reported scores).

This is, of course, exactly the AIEH Skills Passport thesis — and it’s why we built it.

The Junior-Hire Decision in 2026

For a hiring team trying to make a junior hire today, the practical implication is straightforward:

  • A traditional resume packet (resume, cover letter, transcripts) is worth what it has always been worth: ~7 seconds of attention and a predictor not much better than a coin flip.
  • A Skills Passport packet (calibrated 300-850 composite, multi-pillar breakdown, recency indicator, with proof-of-work links alongside) lets a recruiter spend the same 7 seconds and walk away with a much more reliable read.

The signal-to-noise difference compounds over a recruiting funnel. Even a modest improvement in initial screening accuracy carries through to lighter downstream interview load and fewer false-positive offers.

What This Looks Like for the Candidate Side

Candidates who recognize the shift are starting to build their own evidence packets — sometimes ahead of the recruiters asking for them. The Proof-of-Work Hiring Packet — a candidate-side framework for recent grads — has them submitting monthly proof packets: usage receipts for AI tooling, weekly accomplishment reports, GitHub deliverables, resume-as-cover-sheet.

This is a healthy market response. When the supply side starts producing better evidence voluntarily, hiring teams that can intake and evaluate that evidence have a structural advantage over hiring teams stuck on resume-keyword screening.

The Short Version

The entry-level hiring market is broken in both directions. New grads can’t get hired despite credentials. Recruiters can’t distinguish quality candidates despite spending more on screening than they did a few years ago. Both sides are stuck inside an evaluation framework — the traditional resume — that AI has rendered obsolete.

The fix isn’t more interviews or more buzzword filters. It’s better evidence, captured at the front of the funnel, normalized across vendors, owned by the candidate, and surfaced at recruiter time.

That’s the bet behind AIEH’s three workspaces. For recruiters, it’s Hire — calibrated bundles, evidence-based candidate surfacing, no per-test fees. For candidates, it’s Learn — free sample tests, role guides, a Skills Passport that travels. For test providers, it’s Assess — an adapter SDK to reach an installed base.

If your hiring funnel is producing hundreds of applicants and zero confident decisions, the problem isn’t volume. It’s the substrate. Time to switch substrates.


For candidates: If you’re on the supply side of this market — trying to break into tech in 2026 with the deck stacked against juniors — see the candidate-side framework: How to Get Hired in Tech When Entry-Level Hiring Is Down 50% →


Sources