I Remain a Skeptic
blog.jsbarretto.com
August 15, 2026
4 min read
🔥🔥🔥🔥🔥
50/100
Summary
There are concerns regarding the efficacy of large language models (LLMs) for building non-trivial software. Issues such as environmental impact, social and political implications, ecosystem lock-in, and the potential erosion of labor bargaining power contribute to skepticism about their use.
Key Takeaways
- The author does not use large language models (LLMs) for important tasks due to concerns about their efficacy and the lack of substantial improvements in software quality or productivity after four years of development.
- There is a lack of independent studies demonstrating significant productivity gains from AI, with existing research often focusing on irrelevant metrics or being too limited in scope.
- LLM-generated code is often of lower quality than human-written code, with a high proportion of generated pull requests deemed unfit for merging.
- The prevailing philosophy among AI proponents contradicts established software development theories, treating lines of code as a productivity metric rather than recognizing code as an input to the development process.
Community Sentiment
Positives
- Apple, Mozilla, and Firefox have released a record number of bug fixes thanks to AI, proving that software quality has indeed improved.
- Many users report getting more done, faster, thanks to LLMs, showcasing their undeniable benefits when used judiciously.
- Some commenters find LLMs invaluable as debugging aids, accelerating programming tasks and enhancing productivity in tangible ways.
Concerns
- Skeptics argue that the industry has little to show for four years of AI development, questioning the actual improvements in software quality.
- There's a pervasive sentiment that claims of massive productivity gains from LLMs are exaggerated, with many users not seeing observable benefits.
- Concerns arise about the potential for LLMs to create more bugs than they fix, raising doubts about their overall effectiveness.
