跪拜 Guibai
← All articles
Frontend

AI Makes Coding Painless — and That's Exactly Why It's Making You Worse

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

AI coding tools are now fast enough to let a developer ship features without understanding a single line. The downstream cost is a workforce that can glue prompts together but collapses on novel bugs, architecture decisions, or anything the model hasn't seen — a hollowing-out that compounds silently until it's too late to reverse.

Summary

Cognitive science calls it automatic processing: when AI generates code, fixes errors, and scaffolds projects, the brain stays in its comfort zone and retains nothing. The result is technical hollowing out — developers who can't function without the tool. Effortful learning theory and the Feynman Technique point to the same conclusion: friction, struggle, and manual debugging are not obstacles to learning; they are the mechanism of learning.

A four-step training method adapted for the AI era flips the default workflow. Learn concepts from raw documentation first, then write code and troubleshoot completely offline before ever touching an AI tool. The moments of stuttering and confusion are precisely where growth happens — they expose real blind spots that AI otherwise smooths over.

The prescription is blunt: reserve daily no-AI coding time, use AI only for optimization after manual completion, and self-test weekly by explaining technical concepts in plain language without any tools. Leave the resistance for yourself; leave the convenience for the job.

Takeaways
— Code obtained without mental friction is classified by the brain as useless information and forms no lasting technical retention.
— AI is a finished-product output tool, not a training tool — the ability it demonstrates belongs to the model, not to the developer using it.
— Heavy AI dependence degrades four specific capacities: logical thinking, syntax fluency, error-location skill, and low-level conceptual understanding.
— The Feynman Technique adapted for programming means closing all tools and writing core logic, reproducing functionality, and explaining principles from scratch — whatever you can't produce is your real gap.
— Senior engineers avoid AI for learning because core competitiveness is not writing any given piece of code but understanding underlying logic, solving unknown problems, and building systems.
— A sustainable workflow sequences manual effort first (learning, writing, debugging), then uses AI only for optimization and gap-filling afterward.
— Daily 30-minute no-AI coding sessions and weekly oral self-tests on three technical points prevent skill erosion over time.
Conclusions

The argument reframes AI coding tools not as a productivity debate but as a training methodology problem — the same tool that accelerates delivery also eliminates the cognitive load required for skill acquisition, and most developers conflate the two modes.

The fitness analogy is unusually precise: just as muscle grows only through resistance and micro-tears followed by repair, mental models form only through cognitive friction and subsequent reconstruction. AI removes the friction, so the reconstruction never triggers.

The article identifies a workplace sorting mechanism already in motion — developers who can only prompt AI become code porters, while those who can work independently become engineers. This bifurcation is likely to widen as models improve and the temptation to skip manual work grows stronger.

Concepts & terms
Effortful Learning
A cognitive science principle stating that durable learning requires active, resisted mental processing — struggling with material, retrieving from memory, and reconstructing knowledge — rather than passive exposure or easy consumption.
Feynman Technique
A learning method centered on explaining a concept in simple, plain language without reference materials. Gaps in understanding become immediately visible when you cannot produce a clear explanation from scratch.
Technical Hollowing Out
The gradual erosion of a developer's foundational skills — debugging, architecture, syntax fluency — caused by outsourcing all cognitive work to AI tools, leaving the developer unable to function independently.
Source: juejin.cn ↗ Google Translate ↗ Backup ↗