Why the Generation That Uses AI the Most Trusts It the Least
> A behind-the-boos survival crisis at graduation ceremonies, and practical ways for ordinary people to break through in the AI era.
*Note: This is a structured English rendering of a long Chinese forum post. All data and claims originate from the original author's cited sources.*
The Story of Two Commencements
In May 2026, two US graduation ceremonies produced opposite reactions to nearly identical messages:
- University of Central Florida (UCF): Real estate executive Gloria Caulfield told arts, humanities, and media graduates that "the rise of AI will be the next industrial revolution." The crowd booed loudly; someone shouted "AI SUCKS!"
- Carnegie Mellon University (CMU): NVIDIA CEO Jensen Huang said AI is "creating a new industrial age"—and 5,800 graduates applauded.
- Cengage (April 2026): Only 30% of US 2025 graduates found full-time work at graduation, down 11 percentage points from 41% a year earlier.
- Handshake: 60% of the class of 2026 feel pessimistic about job prospects; job postings down 16% year-over-year; applications per opening up 26%.
- Gallup (2025): Only 43% of Americans aged 15-34 think it's a good time to find a job—down 32 points from 75% in 2022.
- Revelio Labs: Entry-level job postings fell 35% year-over-year.
- Stanford AI Index 2026: Employment for software developers aged 22-25 dropped nearly 20% since 2024.
- Recruiter survey (via Stack Overflow blog): 70% of hiring managers believe AI can do an intern's work; 57% trust AI output more than intern or new-graduate output.
- Indeed / Handshake: Industry-wide internships down 11%; tech internships down 30% since 2023, while applications rose 7%.
- Tech layoffs Q1 2026: 78,557 layoffs, with 47.9% of announcements explicitly attributed to AI.
- The New York Times (Aug 2025): One 2023 CS graduate applied to 5,762 tech jobs without a single full-time offer.
- During the British Industrial Revolution, power looms grew from ~2,400 (1813) to 250,000 (1850) while textile workers fell from 240,000 (1820) to 69,000 (1845). Per Oxford Martin School research, output per capita rose 46% between 1780-1840, but real wages rose only 14% while working hours rose 20%.
- Per *The Coming Wave* (Mustafa Suleyman), textile wages were halved in the 45 years after 1770 while food prices soared.
- The Luddite movement (1811-1813) was a rational response—machine-breaking was the only Nash equilibrium when skills were devalued below survival and workers had no capital share in the new wealth. It was crushed by the military; 17 Luddites were hanged in 1813.
- The economists' compensation effect (new tech lowers costs → demand rises → new jobs) holds only long-term. The gap between handloom weavers' displacement and stable new employment lasted at least two generations (40-60 years). The ILO notes destruction comes first; compensation arrives later and slower.
- Why this time differs: the Industrial Revolution replaced physical labor and created lower-skill replacement jobs; AI replaces cognitive labor—exactly the baseline skills young people use to enter the workforce—while the new jobs it creates demand higher skills. Iteration cycles are months, not decades, and the shock is global and simultaneous.
- Cengage, *The Future of Jobs Report 2026*
- Handshake, *2026 Hiring Landscape Report*
- Gallup, *U.S. Job Market Perceptions Survey 2025*
- Stanford HAI, *AI Index Report 2026*
- Revelio Labs, *Entry-Level Job Postings Tracker 2025-2026*
- Stack Overflow Blog, *AI vs Gen Z: How AI Has Changed the Career Pathway for Junior Developers* (2025-12-26)
- Indeed, *Internship Trends Report 2026*
- Challenger, Gray & Christmas, *Tech Layoffs Report Q1 2026*
- The New York Times, *Goodbye, $165,000 Tech Jobs* (2025-08)
- Mustafa Suleyman, *The Coming Wave*
- Oxford Martin School, *Political Machinery*
- ILO, *Technological Changes and Work in the Future*
- Jensen Huang's CMU 2026 commencement address; Business Insider / NBC News / Florida Today / TechTimes coverage of the UCF commencement incident
The difference is the audience: UCF graduates (artists, writers, journalists) are among the first casualties of AI disruption; CMU graduates (computer scientists, engineers) are among its beneficiaries. AI has split the labor market in half—one half on shore, one half drowning.
Key Points: The Data
Key Points: The Engineering Analysis
The author frames the crisis as a system architecture rewrite, not a sentiment problem:
1. The career ladder model broke. The old pipeline—junior staff execute tasks → gain experience → get promoted—assumed execution was the entry point of value. The new model is AI executes ~80% of tasks → humans do QA, tuning, and complex judgment → high-level decisions. Junior execution value fell below the cost of hiring a junior (e.g., a $70k salary vs. $20-200/month AI subscriptions).
2. Internships died for three reasons. AI beats interns on cheap labor; companies can vet talent from GitHub/Kaggle instead; and if basic tasks go to AI, juniors never get the mentoring that transfers skills.
3. A skills-stack cliff. AI flattened the lower rungs (writing code, debugging, designing modules), then demands candidates who are already "senior." Nobody is born senior—seniority requires the rungs AI removed. This creates a deadlock: firms won't hire juniors because AI can do the work; juniors can't gain experience because firms won't hire; without experience they can't prove they "use AI better."
Key Points: History Doesn't Automatically Improve
Key Points: Jensen Huang's "Same Starting Line"
Huang told CMU graduates: "We are all on the same starting line" and "AI is not likely to replace you, but someone using AI better than you might." The author argues this ignores unequal starting points: CMU graduates had four years of Copilot-assisted coding and research; arts graduates' core skills (visual storytelling, creative ideation) can now be approximated in seconds. The advice also presupposes equal opportunity to use AI at work—which requires having a job, which requires experience. Hence: a correct but empty statement in the face of the deadlock. CMU graduates didn't boo because they're already aboard the AI ship—builders, not the optimized.
Key Points: Three Survival Strategies
1. Shift from task-executor to problem-definer. When AI writes the code, value moves to requirements decomposition, boundary judgment (safety, ethics, compliance), quality acceptance, and exception handling—systems thinking plus domain knowledge plus critical judgment. 2. Occupy AI's blind spots. (a) Physical-world complexity: field diagnosis, construction-site coordination, hands-on care; Huang himself highlighted electricians, plumbers, steelworkers, and technicians as AI's trillion-dollar data-center buildout depends on them. (b) Interpersonal trust and emotional labor: reading hesitation in a negotiation, comforting the bereaved—high-value human capabilities outside AI's boundary. 3. Be the AI operator, not the operated. Young people are the first AI-native generation. Cite: Coursera (Jan 2026) data showing developers completing prompt-engineering/AI-collaboration courses saw 58% more career opportunities and ~$18,000 average salary gains. The formula: human judgment × AI execution efficiency = new competitiveness.
Long-term, invest in compounding capabilities: cross-domain translation, system decomposition, meta-learning, and taste/aesthetics—none replaceable by any single model generation.
Conclusion
Fear is rational; avoidance is not an option. Historical losers—the 1780 weavers, 1890 coachmen, 1980 typists—were all "left behind," but each revolution also created winners: those who understood the new rules first. The 2026 rules: tasks vanish but problems don't; execution value falls while problem-definition value rises; single skills depreciate while skill combinations appreciate. The greatest risk is not being replaced by AI, but refusing to understand AI out of fear—and being replaced by those who didn't. The only way to break the negative feedback loop of fear → avoidance → falling behind is to act despite the fear.