01What has already changed
In one year, AI went from an assistant to the main author of code at many companies. A quarter of Y Combinator's Winter 2025 batch had codebases about 95% generated by AI, according to YC's Jared Friedman. In Stack Overflow's 2025 survey, 84% of respondents used or planned to use AI tools, up from 76% the year before.
Collins Dictionary chose "vibe coding" as its 2025 word of the year. Whatever happens next, the way most software starts life has already changed.
02Why trust is falling as use rises
The same Stack Overflow survey found that more developers distrust the accuracy of AI tools (46%) than trust it (33%). The most common frustration was AI output that is almost right but not quite, which is the hardest kind of wrong to catch.
This is the key tension in where vibe coding goes. The tools are good enough to use everywhere and not good enough to trust blindly. That pushes the work toward checking, testing and reviewing rather than writing.
| Kind of software | Likely future | Why |
|---|---|---|
| Prototypes and demos | Mostly vibe coded | Cheap to throw away, fast to test |
| Personal and internal tools | Largely vibe coded, lightly reviewed | Low stakes, high value |
| Marketing websites | Generated, with design decisions explicit | Looks and speed matter more than code |
| Products with user data or payments | AI-written, human-reviewed | Security and liability |
| Safety-critical systems | Slowest to change | Regulation and accountability |
03The skills moving to the center
If AI does the typing, the valuable work moves to either side of it: deciding precisely what to build, and confirming it was built right. Hacker News commenters discussing the future of programming describe engineering shifting toward writing specifications and building validation around what agents produce.
- Specification: describing behavior precisely enough that an AI builds the right thing.
- Verification: tests and checks that prove the code does what the spec says.
- Architecture: structuring systems so AI changes stay contained.
- Judgment: knowing which risks matter for this product and these users.
- Taste: in interfaces especially, deciding what good looks like. Why AI websites all look the same shows what happens when nobody does.
04What happens to programming jobs
The US Bureau of Labor Statistics still projects 10% growth in software developer, QA and tester employment from 2025 to 2035, according to its Occupational Outlook Handbook. Forecasts in either direction carry wide uncertainty right now.
The most consistent view from practitioners is that entry-level work is under the most pressure, while experienced people who can review and take responsibility for AI output become more valuable. Can vibe coding get you a job? and is web design dead? look at what that means for career choices.
05What this means for you
If you build things, learn to vibe code now; it is the fastest way to test ideas. If you want a career in software, learn to specify and verify, because that is where the human value is concentrating. If you need a website, the question has already been mostly answered: describing it to a tool that makes good design decisions beats hand-coding for most small businesses. How do AI website builders work? explains what happens under the hood.
What people who tried it say
Hacker News is split between those who think all coding becomes AI-directed and those who think vibe coding hits a complexity ceiling. Most land on a shared middle: AI does more of the writing, humans keep the responsibility.
- The vibe coder's career path is doomedHacker News
The linked post argues LLM-built projects hit a complexity ceiling; commenters counter that better architecture and context management keep pushing that ceiling up, and debate how much hype shapes both views.
- Ask HN: Is vibe coding a new mandatory job requirement?Hacker News
Commenters report AI coding entering hiring requirements at some companies while arguing that review skills, domain knowledge and knowing what to build matter more.
Looking for more first-hand opinions? Search Reddit for “is vibe coding the future”. We link the live threads rather than summarizing them, so you read them in full.
Our take
"vibe coding" as Karpathy described it, forgetting the code exists, will remain the way prototypes are made. The future of serious software looks more like specification plus AI plus verification, and the people who thrive will be the ones who are good at the first and last parts.
Where this falls short
Predictions about AI and software have a poor track record in both directions. Model capabilities, costs and regulation could all change the picture within a year, and the data here is mostly from 2025.
Sources
- 1.A quarter of startups in YC's current cohort have codebases that are almost entirely AI-generated · TechCrunch
- 2.2025 Developer Survey: AI · Stack Overflow
- 3.Software Developers, Quality Assurance Analysts, and Testers · US Bureau of Labor Statistics
- 4.Replit AI agent snafu: shot across the bow for vibe coding · TechTarget
Researched and written by the Orlume team · Updated October 2026
Still stuck?