For decades the software career path had a reliable first step: a junior role writing boilerplate, fixing routine bugs and running scripted tests while learning the craft. In 2026 that step is disappearing, and the data now comes from multiple independent sources.
Key takeaways:
- Indeed data (via The San Francisco Standard, 20 August 2026): entry-level job postings in San Francisco in June were down 84% on the same period in 2019. Entry-level tech postings have been declining since 2022.
- Stanford HAI 2026 AI Index (13 April 2026), using ADP payroll data: employment of software developers aged 22–25 fell nearly 20% from 2024, while developers aged 30+ at the same firms grew 6–12%.
- The Index also finds AI-related skills now appear in 2.5% of all US job postings, a 297% increase over the decade.
- Where the jobs moved: Indeed Hiring Lab data show machine-learning engineer postings well above their early-2020 baseline while general software-engineer postings are far below it.
- Economists caution that interest rates and the post-2020 hiring correction also contribute; not all of the decline is AI.
What the numbers actually show
The Stanford analysis, led by economists Erik Brynjolfsson, Bharat Chandar and Ruyu Chen, is the more rigorous of the two because it uses payroll records rather than job adverts and compares age groups within the same companies.
| Software developers | Change since 2024 |
|---|---|
| Aged 22–25 | Down nearly 20% |
| Aged 30+ (same firms) | Up 6–12% |
That pattern rules out the simplest explanation. If companies were merely cutting engineering, older developers would be falling too. Instead, firms are keeping experienced staff and not replacing the juniors.
The San Francisco figure is starker but narrower: an 84% fall in entry-level postings in one city, which was also the epicentre of the 2020–22 hiring boom. Indeed Hiring Lab economist Sneha Puri notes the decline began in 2022, before generative AI tools were widespread.
Why juniors specifically
The tasks juniors were hired for are the tasks current AI tools do best: generating boilerplate, writing scripted tests, fixing well-defined bugs, translating a spec into a first draft. The discipline of software engineering is not disappearing; the entry point to it is being automated.
That creates a structural problem the industry has not solved. Senior engineers are made from junior engineers. If the first rung goes, the pipeline for the next decade's seniors goes with it.
Where the demand went
The market is rotating rather than contracting uniformly. Indeed Hiring Lab's indexed data (February 2020 = 100) has shown machine-learning engineer postings well above baseline while general software-engineer postings sit at roughly half of it. The Stanford Index's finding that AI skills appear in 2.5% of all US postings, up almost 300% in a decade, points the same way.
| Role type | Direction since 2020 |
|---|---|
| Machine-learning engineer | Well above baseline |
| All tech postings | Below baseline |
| General software engineer | Roughly half of baseline |
What a learner can do about it
- Skip the tasks AI does. A portfolio of boilerplate CRUD apps no longer signals anything. Build things that show judgement: system design, debugging a real production issue, evaluating AI output.
- Target the growing side. ML engineering, AI infrastructure, data engineering and applied-AI roles are where entry-level demand still exists.
- Get AI-fluent, visibly. The PwC Barometer puts the wage premium for AI skills at 62%. Employers increasingly assume juniors arrive already using these tools well.
- Seek apprenticeship deliberately. Internships, open-source contribution and mentored roles matter more when companies are not creating junior seats.
Concept to take away: task automation versus job automation. AI rarely eliminates a job outright. It eliminates tasks, and if a job was mostly made of those tasks, the job shrinks. Entry-level software work was unusually task-dense, which is why it is the canary. Ask of any role: what fraction of it is codifiable?
Note: balance matters here. Economists quoted across the coverage stress that higher interest rates and a correction from 2020–22 over-hiring also depress junior hiring. The 84% figure is San Francisco-specific; do not generalise it nationally.
