Fewer junior tech jobs? What your first CV must show

Are junior tech jobs really shrinking in 2026?

Junior tech jobs haven't disappeared, but they've become a much thinner slice of what companies hire for. In a report published on 16 October 2026, Anthropic said the junior share of its technical workforce slid from 21% to 14% since early 2025, while the junior share of its technical hires dropped from 27% to 9%. Postings aimed at juniors fell especially sharply after December 2025. Over the same stretch, hours logged by AI agents overtook employee hours in March 2026 and reached roughly six to one by August, and code commits per engineer rose about fivefold in six months. Technical hiring still grew in absolute terms — these are shifting proportions, not layoffs.

Two other numbers from that report matter more for your CV than the headline. Meeting time per employee climbed about 20% year on year, and 16% for technical contributors, which tells you coordination and judgment are now a bigger part of the job than typing code. Roughly a fifth of surveyed employees said their work had shifted from doing the task to managing AI agents. Anthropic itself cautions that one AI company may not represent the wider economy, and the outside evidence is genuinely mixed: Stanford researchers found fewer young hires in AI-exposed occupations, while a Danish study found no entry-level gap between firms that use AI and firms that don't.

The broader market points the same direction without being apocalyptic. Entry-level developer postings in the US have fallen by roughly a fifth since 2025 even as overall tech demand grew, and juniors now make up a smaller share of IT employment than they did three years ago. Meanwhile IBM has said it plans to triple US entry-level hiring in 2026, and firms like Stripe have reshaped new-grad onboarding around reviewing AI-generated code, writing integration tests and pairing on system design. Read those together and the picture is clear: fewer junior slots, redefined around verification and judgment. Your CV has to speak to that version of the role.

What should lead your first CV instead of coursework?

Lead with shipped work. On a first CV, a Projects section sitting directly under a two-line summary beats a Education-first layout every time, because coursework only proves you attended while a shipped project proves you finished something nobody forced you to finish. Put two or three projects above your education, and give each one four things: a plain-English line on what it does, who used it, the stack, and a result. "Built a scheduling tool used by 40 members of my university rowing club; React, FastAPI, Postgres; cut sign-up admin from two hours a week to ten minutes." That's a hiring conversation. "Completed coursework in web development" isn't.

Quality beats volume, and here's the opinion that gets me arguments: three abandoned tutorial clones hurt you more than having no projects section at all. A reviewer who opens your repo and finds a half-finished to-do app with no README learns that you follow instructions and then stop. One finished, documented, deployed thing — with a README that says what it is, how to run it, and what you'd fix next — carries more weight than six starts. If an AI assistant wrote most of it, that's fine, but know where the line sits on what you can honestly claim for an AI-built side project before you write the bullet.

Internships, part-time work and freelance gigs belong in the same tier as projects, not below them. Three weeks of real production access — even on a small team, even unpaid — outranks a semester-long capstone, so write it as experience with a date range. Keep education to the degree, institution, graduation year and at most two genuinely relevant modules or your capstone; drop GPA unless it's outstanding or the employer asks, and drop general-education classes entirely. If you're rebuilding the whole document rather than patching it, a chat-based CV builder will get you to a clean single-column draft faster than wrestling with a template.

How do you quantify impact with no job experience?

You quantify a project the same way you'd quantify a job: measure the thing before, measure it after, and name the tool you measured with. Every junior has numbers available — users, requests, rows, tests, hours saved, latency, build time, bug counts, repo stars, assignment turnaround. The formula that works is action, system, number, outcome: "Cut cold-start time on a Django API from 2.1s to 640ms by adding Redis caching, verified with k6 at 500 virtual users." Nobody asks whether that was paid work. The bullet is specific enough to be interrogated in an interview, which is exactly why screeners trust it.

When you genuinely have no numbers, use scope instead of inventing one. Team size, timeframe, data volume, number of endpoints, number of integrations, test coverage, how many people reviewed your code — these are all countable and all honest. "Shipped 14 endpoints over a six-week sprint with two other contributors, 78% test coverage" says more about readiness than any adjective. Never round a guess up into a statistic; a fabricated "improved performance by 40%" collapses the moment someone asks how you measured it. Most rejected junior CVs fail on this exact axis, which is why duty-list bullets with no numbers read as filler to anyone screening at volume.

Then cut the words that carry no information. "Passionate about clean code", "strong problem-solving skills", "fast learner", "team player" — all of them survive on CVs because they're easy to write and impossible to verify. Replace each with one piece of evidence: a code review you acted on, a flaky test you tracked down, a dependency you upgraded without breaking the build. Aim for every bullet to contain at least one noun a search engine or a parser would index — a framework, a database, a protocol, a metric. If a line would read identically on a stranger's CV, it's taking up space you need for the first screen.

How do you show you can direct AI, not just use it?

Show oversight, not usage. Listing "Claude, ChatGPT, Copilot" in your skills block proves nothing in 2026 — everyone's got that line, and it reads as a tool list rather than a capability. What's scarce, and what the shift inside AI-heavy companies actually rewards, is the ability to review generated output, catch what's wrong with it, and take responsibility for what ships. Write that as work: what you asked the model to do, what you checked, what you rejected, and what the outcome was. "Reviewed AI-generated migration scripts against the staging schema and caught two destructive statements before release" is a judgment bullet, and judgment is what's being graded.

Three bullet patterns do most of the lifting. First, verification: you built the test, the schema check or the review step that caught model errors. Second, orchestration: you broke a task into steps an agent could run, then integrated and shipped the result — say how many steps, how long it took, what you reused. Third, throughput with a quality gate: more output per week without a rise in rejected pull requests or production incidents. If you want the exact wording, there's a fuller set of bullet formulas for AI agent oversight work, including the four numbers worth quantifying and the titles worth matching.

Mirror the employer's vocabulary rather than your own. If the ad says "code review", "evaluation", "prompt engineering", "guardrails" or "human-in-the-loop", those are the strings a parser and a recruiter are both scanning for, and they belong in your bullets where they're true. A skills section is still worth keeping, but group it — languages, frameworks, data stores, cloud, tooling — so the stack is machine-readable rather than one comma-soup paragraph. For the full first-screen checklist aimed at engineers, including how to lay the stack out so it parses, there's a focused guide on getting a software engineer CV through the first screen.

What do screeners actually weigh on a junior CV?

Five things decide a junior CV: relevant experience, the tech stack you can show, impact and ownership, clarity and structure, and whether an ATS can read the file at all. Everything else — the colour accent, the skill bars, the photo, the clever two-column layout — is noise at best and a parsing failure at worst. The order matters too, because a reviewer gives a first-screen CV a few seconds: the top third of page one should carry your strongest project, your stack and one number. If your best evidence sits on page two, assume it was never read.

The formatting rules are blunt and non-negotiable. One column, standard section headings (Experience, Projects, Skills, Education), real text rather than text inside images or boxes, dates in a consistent format, and a PDF exported from a word processor rather than a design tool. Two-column templates are the single most common reason a strong junior CV comes back parsed as nonsense, because many parsers read straight across the page and interleave your two columns into gibberish. Keep it to one page while you're early-career; two pages is fine once you have a couple of real roles, but padding a thin history to fill space is obvious.

Then check the file instead of trusting it. A free instant AI CV analysis gives you a score out of 100, the five category scores above, your strengths and a visual layout analysis — so you can see whether your layout survives extraction and where a screener's eye actually lands before you send it anywhere. Paste the job description in and the analysis lists the skills you're missing against that ad and generates ATS-friendly rewrites in six templates that mirror its keywords. For LinkedIn or Indeed postings, paste the text rather than the URL — those sites block fetching. Fix the structural problems first; wording is the easy part.

Frequently asked questions