From 2010 to 2022, "learn to code" was the default answer to any career question. Laid off? Learn to code. Changing careers? Learn to code. Want a raise? Learn to code.
It was good advice. Software engineering paid well, didn't require a credential, and had more openings than qualified applicants for over a decade. Coding bootcamps boomed. Computer science became the most popular major at most universities.
Then AI arrived. And the bottom fell out.
The Numbers
Entry-level developer job postings dropped 67% between 2022 and 2026. The actual decline in junior hiring was worse. 73%, because postings labeled "junior" were increasingly filled with senior candidates desperate for any role.
The Stanford Digital Economy Study tracked a roughly 20% employment drop for software developers aged 22-25. The Federal Reserve reported a 7.5% unemployment rate for computer engineering graduates. higher than fine arts majors, a statistic that would have sounded like satire in 2020.
The mechanism is straightforward. A senior developer using AI tools produces in hours what a junior would produce in days. The company saves $80,000 in salary plus benefits plus management overhead. The junior never gets hired. The senior gets more done. Everyone involved thinks this is efficient except the junior who can't break into the field and the senior who will have nobody to replace them in five years.
The Career Advice Reversal
The KB compiled X/Twitter consensus on career advice as of May 2026. The shift is the fastest in a generation.
| Advice | Share | Trend | |---|---|---| | Learn a trade (electrician, plumber, HVAC) | 35-40% | Fastest growing | | AI-augmented tech / become AI operator | 25-30% | "Code with AI, not against it" | | Start a business / build an audience | 20% | Creator economy, AI agencies | | Pure "learn to code" (old style) | <10% | Shrinking rapidly | | Just survive / pessimist | 10-15% | Minimize debt, avoid kids, wait |
The single most common piece of advice: "Don't compete with AI. Either use AI as a multiplier in a hard-to-automate domain, learn a real trade that requires physical presence, or build your own business. Pure cognitive entry-level jobs are over."
What's Actually Safe
The careers that survive are the ones with physical presence, legal liability, or licensing requirements. They're the ones AI can supplement but not replace.
Electricians, plumbers, and HVAC technicians are the #1 recommended career on X. The reasoning is simple: AI cannot physically show up and fix your pipes. AI cannot pull wire through a wall. AI cannot pass a local inspection. These careers are recession-proof and automation-proof for the same reason. someone has to be there.
Licensed professionals. tax preparers, nurses, real estate agents, cybersecurity specialists. survive because the state requires a human signature. AI can draft a tax memo but cannot sign a return. AI can suggest treatment but cannot touch a patient. AI can scan for vulnerabilities but cannot sign off on a security audit.
Teachers, therapists, and elderly care workers survive because of human empathy and physical presence. The work requires a person in the room who cares about the outcome, not just the output.
The Junior Loop That Won't Reform
AWS CEO Matt Garman called replacing juniors with AI "one of the dumbest things I've ever heard." Anthropic CEO Dario Amodei said 50% of entry-level jobs may be wiped out by AI. The CEOs disagree on whether this is good or bad. They agree it's happening.
A Harvard study attributed only about 10% of the junior hiring decline directly to AI. The other 57% is traditional cost-cutting dressed as AI rhetoric. Companies are using AI as the excuse to do what they wanted to do anyway: hire fewer people.
The pipeline damage will take years to appear. From 2026 to 2030, companies enjoy the cost savings of hiring fewer juniors. From 2030 to 2035, mid-level talent shortages appear as fewer juniors mature into seniors. After 2035, leadership crisis, institutional knowledge gaps, and salary inflation for the few seniors who remain.
What to Actually Do
The people who survive this aren't the ones who learned the hottest framework. They're the ones who picked a domain where AI is a tool, not a replacement.
An EA credential costs $627 in exam fees and 4-12 months of study. Tax preparation has a massive replacement demand. thousands of preparers retiring, not enough new ones entering. Personal liability means AI can't replace the signer. Client trust means AI can't replace the relationship. The math is simple: if you're choosing between a credential that AI is actively displacing and one that AI structurally can't displace, pick the second one.
---
*Sources: ByteIota, Stanford Digital Economy Lab, Federal Reserve, Anthropic, AWS, Harvard Business School. X consensus analysis from KB discourse compilation (May 2026).*