Who will train the next senior developer?

1. AI became the default, fast

Generative coding assistants showed up in late 2022. By 2025, using them (or planning to) wasn’t really a choice anymore; it was the baseline. Adoption jumped from 70% to 84% in two years, even as developers’ trust in the output went the other way: favourable sentiment slid from over 70% to about 60%.

Figure 1. Share of developers using or planning to use AI tools in their workflow. Source: Stack Overflow Developer Survey (2023–2025).

A bar for each year rises from 70% (2023) to 76% (2024) to 84% (2025), showing AI tools becoming near-universal among developers within two years.

2. Employers now expect AI too

Developers aren’t the only ones reaching for AI. Employers are writing it into the job description itself. The share of ads asking for AI skills has roughly tripled since 2019, and the steep part of the climb starts right after ChatGPT launched. In Australia it hit about 6.9% of all postings by 2026; in the UK, over 8%.

Figure 2. Share of job postings mentioning generative-AI skills, monthly. Source: Indeed Hiring Lab AI-at-Work Tracker (2026), CC BY 4.0.

Three rising lines (Australia, US, UK). All are flat near 2% until late 2022, then climb steeply, reaching roughly 5–8% by 2026 - AI demand entering hiring itself.

3. Meanwhile, the jobs that make developers are vanishing

Here’s the squeeze. Software development postings boomed through the pandemic, then fell off a cliff. Australia is down 56% from its 2022 peak; the US is down 68%, which puts it below where it sat before COVID. Now look at nursing, a job AI can’t do: barely a scratch. The damage lands squarely on the knowledge work AI has got good at.

Figure 3. Job postings index by sector, Australia and US, monthly (Feb 2020 = 100). Solid = Software Development; dotted = Nursing. Source: Indeed Hiring Lab Job Postings Tracker (2026), CC BY 4.0.

Software-development postings (solid lines) spike to 230–270 in 2022 then crash to ~119 (AU) and ~73 (US) by 2026. Nursing postings (dotted) stay high throughout, showing the fall is specific to AI-exposed tech work.

4. And the door to the bottom rung is closing

Where firms are still hiring developers, they want experience a junior can’t have yet. Between 2022 and 2025 the share of tech ads demanding five or more years climbed from 37% to 42%, while the two-to-four-year slice shrank. Stanford reckons employment among the youngest developers, aged 22 to 25, has already dropped about 20% from its 2022 peak. The bottom rung is disappearing.

Figure 4. Experience demanded in tech job postings, Q2 2022 vs Q2 2025. Source: Indeed Hiring Lab (2025); youth-employment figure from Stanford Digital Economy Lab (2025).

A slope chart: the red line (5+ years required) rises 37%→42%, the blue line (2–4 years) falls 46%→40%, and entry-level (0–1 year, grey) is flat. Employers increasingly require seniority.

5. So who is senior in 2035?

Seniors aren’t born, they’re grown, usually over six to eight years, out of juniors. The ones we have now were hired before the 2022 crash, so the shelves still look full. But follow the cohorts forward. How many people cross into “senior” in a given year depends on how many juniors got hired six years earlier. Freeze junior hiring now and the shortfall isn’t some distant risk; it’s already on the books for the 2030s.

Figure 5. Illustrative scenario model - not a forecast. The annual inflow of new senior developers is modelled as junior intake lagged six years; intake collapse anchored to Indeed (2026) and Stanford (2025) figures. Assumptions are stated in the code. Source: author’s model on cited data.

A projection from 2024 to 2035. The “needed” line rises gently. Both scenarios sit at 100 until ~2028, then the “paused hiring” line (red) falls toward 55 while the “rehire now” line (blue) recovers toward 100 - the shaded red gap is the looming senior shortfall.

Data & method

All five charts are built in R, with tidyverse and plotly. The job postings and AI-skill series come straight from Indeed Hiring Lab’s public repositories (CC BY 4.0) through prep_data.R; the survey and experience figures are typed in from the cited reports. Chart 5 is a clearly labelled scenario model, not a forecast, and its assumptions are right there in the code. The postings numbers are an index (February 2020 = 100), not raw counts, so you can compare countries but shouldn’t read absolute volumes into them. One honest caveat: the pandemic boom and bust hit every sector. What’s specific to developers is the failure to recover, and the rising experience bar, and both of those land after generative AI arrived.

Acknowledgements

I acknowledge the use of generative AI tool to help debug the R visualisation code. All data were independently verified against the original sources, all code was executed and checked by me, and the final analysis, design decisions and interpretation are my own. Data: Indeed Hiring Lab (CC BY 4.0). Brand colours follow The Conversation style guide.