No Human Will Ever Write Code Better Than LLMs? Better at What?
Post · https://dev.to/dannwaneri/no-human-will-ever-write-code-better-than-llms-better-at-what-1dme
A tweet went around saying: "No human will ever write code better than LLMs. Accept this fact, stop...
First seen by Oriel on .
AI summary
This article argues that LLMs may produce code quickly, but human expertise remains superior at judging correctness, maintainability, edge cases, and whether code should be built at all.
- Lines per minute measures typing, not software quality; large amounts of incorrect or fragile code can be worse than no code.
- LLMs can generate plausible code that mishandles real constraints, such as retrying payment requests without idempotency and silently converting failures into successful zero-dollar charges.
- Coding benchmarks reward passing tests but often fail to penalize poor design, encouraging techniques like unnecessary try/catch blocks or type casts.
- The author presents a hierarchy of “better”: productivity, correctness, maintainability, and finally the wisdom to avoid unnecessary implementation.
- Maintainable software is readable, simple, adaptable, and resistant to breaking when requirements change.
#llms #software engineering #code quality #maintainability
Generated by gpt-5.6-sol. Check the original for details.
Metric trends
Last 30 days · Up to 120 recorded samples per source and metric, with reposts kept separate · First and latest retained
Reactions
Total public reactions reported by the source; distinct from points and likes.
- DEV Community: 70, latest observation ; +54 over observed period since
Comments
Comment count reported by the source.
- DEV Community: 57, latest observation ; +50 over observed period since
Aliases
| Source | External ID | URL |
|---|---|---|
| DEV Community | 4919957 | https://dev.to/dannwaneri/no-human-will-ever-write-code-better-than-llms-better-at-what-1dme Read discussion |