HN
Today

So where are all the AI apps?

Claims of AI boosting developer productivity by 2x-100x are commonplace, but this analysis of PyPI data finds little evidence of a broad software boom. While new AI-focused packages see accelerated updates, overall package creation and general update frequency remain largely flat. This challenges popular narratives, suggesting AI's impact on software output is concentrated and driven by hype rather than universal productivity gains.

5
Score
1
Comments
#10
Highest Rank
6h
on Front Page
First Seen
Mar 24, 2:00 PM
Last Seen
Mar 24, 7:00 PM
Rank Over Time
101018212021

The Lowdown

This article scrutinizes the widely-held belief that AI significantly boosts developer productivity, using empirical data from PyPI, the Python Package Index. Proponents often claim massive productivity gains, leading to the expectation of a surge in new software development. However, the authors investigate whether this 'AI effect' is actually visible in real-world software repositories.Key findings from their analysis include: No discernible increase in the overall rate of new package creation on PyPI since the advent of ChatGPT. Similarly, the general rate of package updates across the entire ecosystem shows only a marginal increase, and any modest upward trend largely predates modern AI coding tools, potentially attributable to factors like continuous integration adoption. A notable exception is observed in packages specifically about AI: newly created and popular AI-related packages demonstrate a significant (>2x) increase in their update frequency. This rapid iteration is not seen in non-AI packages, even popular ones. The effect is highly concentrated within the most popular AI packages, suggesting it's not merely a general 'AI skill' boost for developers, nor is it widespread across all software. In conclusion, the data suggests that AI's primary measurable impact on the PyPI ecosystem is not a widespread revolution in software productivity. Instead, it manifests as an intense, focused burst of development and iteration within the AI ecosystem itself, potentially fueled more by significant funding and hype than by a universal enhancement of developer capabilities.