跪拜 Guibai
← All articles
Frontend · Backend · AI Programming

AI Is Eroding Pure Code Skills; Engineers Need Psychology and Product Sense Now

By XCaptain ·
Read original on juejin.cn ↗ Google Translate ↗ Alt translation

Competing on technical implementation alone leaves engineers with diminishing leverage as AI absorbs more of that work. The durable edge is human judgment across product, psychology, and business, which is what AI won't automate.

Summary

Every model release shaves worth off purely technical engineering skills. A user story fed as a prompt now beats a rigorous technical spec for many tasks, so the daily job looks less like writing code and more like managing AI as a product owner. The once-formidable technical moat has dried up, and the author says the skills that matter are shifting from technical to human.

The post moves from frustration to strategy: building SaaS used to be a high-barrier problem worth loving, but now it's no longer hard. The author argues engineers must think at a larger scale — national, global, even cosmic — because a 100- or 1,000-person engineering team's productivity is now available at negligible cost. That means expanding beyond code into product, marketing, and sales, and learning things AI can't help with: product positioning, creativity, psychology, and how to delight people. These are human problems with no standard answer or test set, and pattern-matching AI can't touch them.

Takeaways
— Every new AI model release reduces the relative value of purely technical skills like coding and framework knowledge.
— For many tasks, a user story written as a prompt outperforms a detailed technical specification.
— Building enterprise software today is easier than building a simple website five years ago.
— The author's strategy is to learn product positioning, creativity, psychology, marketing, and how to delight humans, because AI can't solve those open-ended problems.
— AI excels at pattern matching but cannot comprehend what it means to be human.
Conclusions

A user story outperforming a technical spec suggests LLMs have internalized much of standard engineering knowledge, leaving the bottleneck at articulating intent and constraints.

The author's list of human skills — psychology, marketing, delight — is framed as AI-proof, but those fields are already being reshaped by AI; the real moat may be the ability to judge what people actually want, not the skill itself.

Source: juejin.cn ↗ Google Translate ↗ Backup ↗