> ## Content Index
> Fetch the complete content index at: https://www.stirup.si/llms.txt
> Use this file to discover other available public pages before exploring further.

# AI Tools Boost Developer Activity But Not Software Releases
- URL: https://www.stirup.si/ai-tools-boost-developer-activity-but-not-software-releases-90a72ea7/
- Published: 2026-10-13T09:33:59.000Z
- Updated: 2026-10-13T09:33:59.000Z
- Description: A 2026 study found AI coding tools increased developer activity by up to 180 percent, but actual software releases only rose by 30 percent, highlighting a growing gap between code production and deployment.
- Author: Satish Das
- Tags: Development, News, Stirup Research

## The Impact of AI on Developer Productivity and Output

A 2026 study by the National Bureau of Economic Research analyzed over 100,000 GitHub developers and found that the introduction of AI coding tools significantly increased coding activity. Autocomplete features boosted activity by about 40 percent, while interactive coding agents increased it by approximately 140 percent, and autonomous agents by 180 percent. However, this surge in commits did not translate proportionally into software releases. The same study noted that an 180 percent increase in commits led to only a 30 percent increase in actual releases, indicating a growing gap between code production and deployment. This suggests that while AI tools may enhance productivity, the value of expert judgment and oversight in software engineering remains critical. The ratio of commits to releases increased by a factor of roughly two and a sixth, meaning each release now involves significantly more generated code than before.

## The Role of Expertise in Software Engineering Amid AI Advancements

The role of expertise in software engineering is evolving as AI tools become more capable of generating code. While AI can produce functional syntax quickly, the deeper aspects of software engineering—such as system design, architecture, and problem-solving—remain the domain of human experts. As noted in *The Attractive, Incomplete Idea*, there is a growing belief that software engineering has become easier due to AI, but this overlooks the complexity involved in creating robust, scalable systems. The ability to write code is no longer the primary differentiator; instead, expertise in understanding user needs, managing technical debt, and ensuring security and performance remains critical. As *What Software Engineering Actually Consists of* explains, the discipline is more about solving complex problems than merely writing lines of code. Thus, while AI may lower the barrier to entry for certain tasks, the value of deep expertise in software engineering continues to grow.

## Economic Shifts in Software Businesses and Investment Trends

As code becomes increasingly cheaper to produce, the economic dynamics of software businesses are shifting. According to Part IV of the book, competitive advantage is migrating from the codebase itself to the broader system surrounding it. This shift is reflected in investment trends, where capital is being directed toward areas beyond traditional hardware-heavy categories. Research indicates that in mature organizations, code is rarely the most expensive output of the engineering process. For instance, a near-tripling of commit volume across 100,000 developers represents a significant productivity shift in the industry. However, this increase in commits does not always translate to a proportional rise in releases, with the ratio of commits to releases increasing by a factor of roughly two and a sixth. This suggests that while code generation has accelerated, the pace of releases has not kept up, highlighting the growing complexity and integration required in modern software development.

## The Limitations and Quality of Evidence in AI and Software Engineering Studies

The limitations and quality of evidence in AI and software engineering studies are often overlooked in public discourse. Appendix B highlights that many studies are vendor-sponsored, have small sample sizes, or are based on findings that have been partially superseded by newer tooling. Additionally, some authors explicitly caution against over-interpreting their results as causal relationships. These factors complicate the reliability of conclusions drawn from such studies. The distinction between functioning and dependable software remains critical, as the gap between them represents the core economic value of engineering. As code production increases, the ratio of commits to releases can rise significantly, indicating that more code does not necessarily equate to more value. This underscores the importance of expertise in ensuring quality and reliability, even as the volume of code grows.

## Sources and reporting notes

- [FreeCodeCamp](https://www.freecodecamp.org/news/the-state-of-ai-and-software-engineering-book/?ref=stirup.si)

Based on published source material. No independent hands-on testing is claimed.