Follow ZDNET: Add us as a preferred source on Google.
ZDNET’s key takeaways
- AI-generated code requires stepped-up human oversight.
- Experts advise keeping AI-generated code in a sandbox.
- At best, AI may do about 80% of the work in building software.
We’ve been hearing incessantly how AI tools and vibe coding mean less need for human coders and programmers. Maybe it’s time to rethink the logic of that argument.
More human oversight
AI — and all that associated vibe stuff — is not diminishing the importance of human coders. If anything, AI requires even more human oversight when it comes to generating and implementing software, argued Michael Li in a recent article in Harvard Business Review.
Also: I took Harvard’s free online coding classes to better catch AI’s errors – and they’re legit
Such tools make coding experience “more — not less — important,” Li said. AI cannot replace real software engineers and coders. He pointed to a recent study that suggested “that while developers estimated that AI made them 20% faster, it actually made them 19% slower.”
When it comes to software design, creation, and implementation, it goes well beyond simply generating code. “Make sure every change it makes is double-checked — with automatic checks, simple tests that confirm things still work, and at least one human review,” said Li, founder and CEO of The Data Incubator and president of Pragmatic Institute.
Keep it in a sandbox
At this point, keep AI-generated development in a sandbox, Li advised. “Never give it the keys to live customer data, and routinely check for basic security mistakes like files or storage left open to the public. Keep experienced engineers in charge of the design, the rules, and the safety checks so AI’s speed doesn’t turn into costly failures.”
Also: Worried about AI coding? Why the invention of power tools is the blueprint for your career future
There are many voices agreeing with Li’s premise that AI-generated software development is not an existential threat to software jobs at this time. Saying AI will “replace software engineers misses the bigger picture,” said Christel Buchanan, founder of ChatandBuild. “Execution is getting cheaper. Direction, judgment, and creativity are becoming more valuable.”
At best, AI may do about 80% of the work in building software, Buchanan explained. “But that last 20% — defining edge cases, architecting for scale, shipping with intent — that still requires a human mind. I don’t think AI is replacing engineers. It’s reshaping the job into something more strategic, more product-minded, and honestly, more fun.”
AI will scale sloppiness
The greatest risk with leaving code production to AI complacency, said Alok Kumar, co-founder and CEO of Cozmo AI, is this: “If your processes are sloppy, AI will scale that sloppiness.”
The advantage AI brings to the table is that it “compresses feedback loops and allows engineers to focus on problem solving rather than mechanical tasks,” Kumar said. “Treat it not as a replacement, but as a true 10x value addition to human engineers.”
Software engineers and programmers should elevate their roles where human judgment adds distinctive value, said Tanner Burson, an engineering leader at Prismatic.
These areas include “system architecture, critical decision-making, production debugging, and staying connected to users’ needs,” Burson said. “The most complex reasoning, nuanced logic, and abstract thinking that development requires will remain challenging for AI systems.”
Also: Vibe coding feels magical, but it can sink your business fast – here’s how
“The challenge is to thoughtfully integrate AI capabilities to enhance developers’ productivity while maintaining a human-centered approach to solving customers’ real problems,” said Burson.
Such expectations need to be level-set in line with what is still the relative immaturity of AI code output.
In his HBR report, Li pointed to the experience of Jason Lemkin, startup founder, VC, and tech blogger, who live-tweeted his AI coding journey “with infectious enthusiasm, riding the wave of possibility that vibe-coding promised — the dream that anyone could build software through natural language alone, freed from the tedium and rigors of traditional engineering.”
Within a week, Lemkin’s experiment flopped. “The AI agent had caused a catastrophic failure: it had gone rogue and wiped his production database entirely, despite explicit instructions to freeze all code modifications. The speed and apparent ease of AI-generated code had seduced builders into abandoning the very guardrails that prevent such disasters.”
We need to adapt
The lesson learned is that AI-generated code “demands more rigorous verification, not less,” said Li. “We need to adapt to a fundamentally different way of writing code. The future likely involves collaboration between human engineers and AI tools, with humans providing architectural vision, rigorous testing, and securing infrastructure while AI accelerates implementation tasks.”
Comments are closed, but trackbacks and pingbacks are open.