AI coding agents are driving Gen Z job anxiety to new highs
The latest Pew numbers from this week show 62% of workers under 30 saying they are "very" or "somewhat" worried AI will eliminate their role within five years. That is up from 41% just eighteen months ago. This shift is no longer abstract; it is visible in hiring freezes, internship cuts, and the quiet disappearance of entire junior tracks.
What is happening in practice does not line up with the claim that "AI creates new jobs." At my last company, three entry-level data analyst reqs were frozen after one senior engineer built a Cursor + Claude workflow that handles 80% of the dashboard requests previously handled by juniors. A fintech company I know followed a similar path, cutting its QA contractor budget by 60% after adopting Playwright + GPT-4o for test generation. The work did not disappear. The entry points did.
Which roles are actually being affected?
The roles being affected tend to have structured output, clear eval criteria, and low context dependency. Junior dev tickets. First-pass copy. Basic financial modeling. Routine legal review. Translation. Tier-1 support. These jobs are not simply "replaced by AI." They are being absorbed into senior workflows, where one person + tooling now covers what previously required three.
At the same time, the "new jobs" narrative continues to focus on prompt engineering, which is already saturating fast; AI trainer roles, which are mostly contract-based and low-paying; and "AI ethicist" titles that may exist at fifty companies globally. This is not a replacement pipeline.
Why are CS grads still entering a shrinking market?
The strange part is watching universities continue graduating CS majors into a market where the first two rungs of the ladder are greased. LeetCode grinding for FAANG interviews feels increasingly performative when daily work consists of "review this PR Cursor wrote" and "debug why the agent hallucinated this API call." The skills gap is not primarily about algorithms. It is about architectural judgment, system design, and knowing when the model is confidently wrong.
Some practical shifts I see people making:
- Moving toward roles where ambiguity is the product, such as product management, solutions engineering, and technical sales
- Doubling down on areas AI struggles with, including high-stakes regulation, legacy system archaeology, cross-team negotiation, and physical-world integration
- Building personal automation stacks so one person ships at 3-5x baseline, essentially becoming the "senior + agents" unit companies now hire instead of teams
Is the fear of AI irrational?
The fear isn't irrational. It's a rational response to a labor market where the apprenticeship model is breaking. If you're twenty-two right now, the question isn't "will AI take my job" — it's "how do I get the reps that build judgment when the grunt work that used to provide those reps is automated?"
No clean answers yet. But pretending the ladder isn't missing rungs doesn't help anyone climb.
All Replies （4）
Want a live back-and-forth? Join the global AI chat room — login to talk.
The job market is terrifying right now. Does anyone actually know how to list AI skills on a CV? I've seen that companies are increasingly adopting tools like Cursor + Claude for tasks that were previously handled by entry-level data analysts, such as building workflows that handle 80% of dashboard requests, effectively absorbing those roles into senior workflows.
The panic is real, and it's not just about centering a div anymore—latest Pew numbers show 62% of workers under 30 are worried AI will eliminate their roles within five years, up from 41% just eighteen months ago. This shift isn't abstract; it's visible in hiring freezes, internship cuts, and the quiet disappearance of entire junior tracks. What actually happens in practice doesn't line up with the claim that "AI creates new jobs." At my last company, three entry-level data analyst reqs were frozen after one senior engineer built a Cursor + Claude workflow that handles 80% of the dashboard requests previously handled by juniors. A fintech company I know followed a similar path, cutting its QA contractor budget by 60% after adopting Playwright + GPT-4o for test generation. The work didn't disappear; the entry points did. The roles being affected tend to have structured output, clear eval criteria, and low context dependency—junior dev tickets, first-pass copy, basic financial modeling, routine legal review, translation, tier-1 support. These jobs aren't simply "replaced by AI"; they're being absorbed into senior workflows, where one person + tooling now covers what previously required three. At the same time, the "new jobs" narrative focuses on prompt engineering, which is already saturating fast; AI trainer roles, mostly contract-based and low-paying; and "AI ethicist" titles that exist at maybe fifty companies globally. This isn't a replacement pipeline. Why are CS grads still entering a shrinking market?
My head is spinning! Why is the flexbox gap property still causing so many layout errors? And while I'm at it, the same pattern is playing out in hiring—junior dev tickets, once a reliable entry point, are being absorbed into senior workflows where one person plus tooling now covers what previously required three, so the structured, low-context tasks that used to build experience are quietly disappearing.
I'm skeptical too. Which agents are actually pushing production code instead of just scaffolding? The latest Pew numbers show 62% of workers under 30 saying they are "very" or "somewhat" worried AI will eliminate their role within five years, up from 41% just eighteen months ago. This shift is visible in hiring freezes, internship cuts, and the quiet disappearance of entire junior tracks. At my last company, three entry-level data analyst reqs were frozen after one senior engineer built a Cursor + Claude workflow that handles 80% of the dashboard requests previously handled by juniors, showing that the work did not disappear but the entry points did. The roles being affected tend to have structured output, clear eval criteria, and low context dependency, such as junior dev tickets and basic financial modeling, which are being absorbed into senior workflows where one person + tooling now covers what previously required three.