When AI changes the work · 5 min read
Junior developer in 2026: how do you get experience when jobs require experience?
Software development is changing. For new developers, a small shipped project with real users offers better evidence than another tutorial alone.
Written by Nivaan, founder of OffLadder · Last reviewed · Sources listed at the foot of this page
The short answer
The difficult part of becoming a developer is not learning how to ask an assistant for code. It is learning how to recognise what is broken, decide what should exist and maintain something another person depends on. Those skills usually come from doing the work. Junior hiring is where that loop is under pressure.
Separate headlines from measurements
Stanford researchers reported in their August 2026 revision that employment among 22-25-year-olds in highly AI-exposed US occupations was 19% below the path implied by less-exposed peer occupations. They also described important caveats and did not claim to have proved AI caused every part of the difference. Indeed, meanwhile, reported a rebound in US software job postings concentrated in senior and AI-related roles. Together the studies suggest a market with opportunities and a difficult entry point, not the extinction of software engineering.
Make evidence that resembles experience
A tutorial demonstrates that you followed one person's path through a known problem. A useful tool demonstrates that you listened to a user, chose trade-offs, managed errors, handled change and shipped something. It need not be impressive. A working rota organiser for a volunteer group beats a complex demo nobody uses, provided you can explain the decisions.
Use AI coding assistance openly. Keep a record of where it accelerated the project and where it misled you. Employers who care about judgement will want to know both. Do not pretend assisted code came entirely from you.
Stop trying to look experienced. Start doing work that produces experience
A polished clone of a popular app proves you can follow a familiar pattern. A tiny tool that solves a specific annoyance for one actual user can prove more: you had to understand the problem, make trade-offs, test failures and support the thing after it shipped. Those are the habits hiring teams struggle to assess through credentials alone.
For example, build an appointment-reminder tool for a local club, with permission. Interview the organiser about the current process. Pick the smallest version: a form, a reminder and an easy correction path. Consider accessibility, data retention and what happens when the reminder does not arrive. Avoid collecting sensitive information unnecessarily. Then ask the organiser to use it for a week. The interesting part of the portfolio is what broke and what you changed, not that an AI coding assistant generated the initial code.
Keep a failure log that demonstrates engineering judgement
For each project, record the requirement, first approach, test, failure and revision. A failure log could read: 'Generated code accepted malformed dates; added server-side validation and tested boundary cases.' Or: 'The interface looked fine with a mouse, but the submit control was unreachable by keyboard; changed focus order and retested.' You are showing how you evaluate outputs, not pretending the assistant never helped.
Add a short security check: secrets outside source control, minimal permissions, no personal data in prompts, input validation and simple error handling. If the application handles real money or sensitive data, an informal portfolio project is not a substitute for professional review. Demonstrate care in the boundaries you set.
Three entry routes worth comparing
Conventional junior engineering still exists, but requirements and applicant volumes vary. Technical implementation or support engineering can use debugging and communication, sometimes with a different coding mix. Accessibility or QA engineering rewards systematic testing and can be a bridge toward broader engineering. Do not assume any is an easy back door. Read a dozen local listings and tally required skills, experience and salary ranges before treating it as a route.
If your portfolio project leaves you energised by debugging, software engineering may still fit. If you liked gathering requirements but dreaded the code, investigate product operations or solutions roles. If the most satisfying part was finding edge cases, investigate testing and quality. A career question becomes much easier once the answer has an object attached to it.
Where OffLadder fits
OffLadder is not a substitute for code practice or a real technical mentor. It can help you test adjacent directions and keep track of what happens when you do. Ask Explore for options that use technical problem-solving, run an experiment with an actual user, then tell Laddie which tasks pulled you in. Start at https://offladder.com/. The answer may be to become a better developer, or to take your engineering instincts somewhere less obvious.
Questions people ask
- Is computer science still worth studying?
- It can be, depending on your goals, cost and location. Combine theory with real projects and people rather than relying on the qualification alone.
- Should I specialise in AI?
- First discover what kind of technical work you like: systems, interfaces, testing, data, hardware or deployment. 'AI' is too broad a substitute for that choice.
Directions you could test
Sources
- Canaries in the Coal Mine? Six Facts, revised August 2026 · Stanford Digital Economy Lab
- AI and Job Postings: From Destruction to Creation? 8 July 2026 · Indeed Hiring Lab
How this page is written
We only state what can be checked. Where a page makes a claim about work, automation or the labour market, it links to a primary source rather than paraphrasing a secondary summary. We do not publish predictions dressed as facts, invented statistics, or lists of AI-proof jobs. Pages are reviewed and dated when the argument or the underlying evidence changes. More about OffLadder.