Vibe coding and agentic engineering are getting closer than I'd like
simonwillison.net
May 6, 2026
8 min read
🔥🔥🔥🔥🔥
53/100
Summary
Vibe coding and agentic engineering are increasingly converging in AI development. Joseph Ruscio discussed these trends in the context of AI coding tools on Heavybit’s High Leverage podcast.
Key Takeaways
- Vibe coding and agentic engineering are increasingly overlapping in software development practices.
- Vibe coding is characterized by a lack of concern for code quality, while agentic engineering involves professional software engineering principles such as security and maintainability.
- The author expresses guilt over not reviewing code generated by AI tools, questioning the responsibility of using unreviewed code in production environments.
- The reliance on AI coding tools has increased the author's ability to tackle more complex challenges, but it raises concerns about the quality and reliability of the output.
Community Sentiment
Positives
- LLMs can significantly enhance coding productivity, allowing developers to produce thousands of lines of code quickly, which could lead to innovative applications and solutions.
- With proper training and verification, LLM-generated code can be of higher quality than human-written code, suggesting that AI can improve software engineering standards.
- The ability to quickly generate code allows for rapid prototyping and experimentation, which can accelerate the development lifecycle and foster creativity in engineering.
Concerns
- The overwhelming volume of LLM-generated code may lead to a chaotic software landscape, making it difficult to maintain or find quality code, which raises concerns about long-term sustainability.
- There is a risk that reliance on LLMs could exacerbate existing issues in software engineering practices, leading to undisciplined coding and architectural decisions that are harder to review.
- The subtle errors introduced by LLMs can create more complex review challenges, as code that compiles may still harbor hidden bugs or architectural flaws, increasing the cognitive burden on engineers.
