AI Boom Fuels Modest Software Projects, Not Blockbusters

Overview

If artificial intelligence is as transformative as its proponents claim, a logical consequence should be an explosion of new, groundbreaking software products. Yet, despite the widespread adoption of coding agents and automated development tools, many observers are puzzled by the lack of blockbuster applications emerging from this technological shift. Where is all this promised software?

This question has circulated in tech circles for years, but the answer appears to be less dramatic than expected. Rather than creating singular, world-changing platforms, AI is fueling a distributed boom in modestly successful software projects. The data suggests that while the volume of new software has increased significantly, the growth is concentrated in hobbyist and niche tools rather than in massively popular applications.

"As far as I can tell, the answer is that it’s spread out over a lot of modestly successful projects."

Analysis of various software marketplaces indicates that somewhere between two times and twenty times more new software is being created, depending on the metric used. However, there is no corresponding growth in wildly successful projects outside the AI toolchain itself. Instead, there is approximately a 2x growth in software that achieves moderate success, serving a few hundred to a few thousand users.

Examining the Data

To understand where this new software is appearing, an analysis was conducted across approximately twenty separate sources. These included software marketplaces, plugin registries, and modding platforms that provide transparent installation counts. The study compared the number of successful projects between the periods of 2022-2024 and 2025-2026, effectively measuring the impact after agentic coding hit the mainstream in mid-2025.

The results show a clear jump in projects that are moderately, rather than wildly, successful. Specific examples from the data include:

• NVDA Screen Reader Addons: The directory saw slightly more than twice as many new addons with over a thousand downloads, though there was no change in addons reaching the ten-thousand-download threshold.
• HACS Community Plugins: Similarly, the open-source home automation store reported double the number of new addons with over a hundred installs, with no growth in those exceeding one thousand installs.
• Steam Games: New Steam releases garnering over a hundred reviews are now approximately 60% more common than they were before 2024.
• Other Platforms: Consistent patterns were observed in Obsidian plugins, Firefox and Chrome extensions, Nexus game mods, and Blender extensions.

While some major platforms like the Google Play store and WordPress plugins did not exhibit this specific pattern, the aggregate evidence strongly suggests that AI is driving a surge in hobbyist software development.

Why Focus on Addons and Plugins?

Focusing on addons, mods, and plugins provides a clearer picture of AI’s actual impact on software utility for a few key reasons. First, these platforms typically offer reliable installation metrics, allowing for an objective assessment of usage. While the explosion in AI-generated repositories is visible on platforms like GitHub, it is difficult to gauge real-world adoption without checking how many people are actually using the code.

Second, examining these smaller projects helps filter out the noise of selling shovels. Pointing to successful software tools that help people code with AI does not necessarily indicate that AI is solving ordinary problems for end-users. It merely reflects the commercialization of the AI development workflow itself.

AI-generated game mods and ports serve as a prime example of genuine utility. Gamers using these tools are not attempting to revolutionize the world with AI coding; they simply want to play multiplayer versions of classic games like Halo CE in a browser, fly planes in Helldivers, or experience Super Smash Bros. with rollback netcode online. The data indicates that AI is effectively satisfying these specific, practical needs.

The Nature of the Boom

A critical question remains: were these projects actually generated by AI? Response rates on this topic are low because hobbyist communities often view AI involvement negatively. In rare instances where developers admit to using AI, they frequently face backlash. Consequently, voluntary disclosure data is unreliable, as the stigma attached to AI tagging discourages honest reporting.

However, the correlation is strong. A 2x growth in successful projects between the pre- and post-agentic coding eras is significant, especially when viewed against a backdrop of 10x growth in total project creation. While not every new project involves AI, the timing and scale suggest that AI is a primary driver of this boom. When critics ask why AI has not improved software, they are often looking for a specific trend of high-impact applications. The trend exists, but it is directed toward small-scale projects that benefit niche communities rather than mass-market audiences.

Why Small Projects and Not Blockbusters?

The concentration of growth in moderately successful projects, rather than dominant ones, can be attributed to several structural factors. Large, successful software companies often already possess all the visible features their user base desires. There is little demand for radical new features in mature products like macOS or Gmail because the core functionality is already optimized for the majority of users.

Furthermore, large enterprises are often technically paralyzed by their existing feature sets. Adding new capabilities to sprawling, legacy codebases is non-trivial, even for highly capable coding agents. Additionally, many improvements in major software are subtle, focusing on enterprise compatibility or internal reliability systems that end-users rarely encounter directly.

It remains unclear why there are not more wildly successful hobbyist projects. It may simply be difficult to generate a truly exceptional product idea, a skill that AI cannot currently replicate. Alternatively, it may take years for current thousand-user tools to mature into hundred-thousand-user phenomena. The current crop of AI-assisted projects may just need time to develop a lasting user base.

Methodology and Transparency

This analysis was conducted with the assistance of AI tools. The researcher provided extensive direction on where and how to investigate the data, but acknowledged that scraping and analyzing such a large volume of information would have been unfeasible manually. The underlying research data has been made publicly available for further scrutiny or agent-based analysis.

The findings suggest that the AI boom in software is real, but it is democratizing the creation of useful, niche tools rather than inventing the next generation of global super-apps. For now, the dividends of agentic coding are being felt in the plugins, mods, and extensions that power the everyday hobbies of millions, rather than in the software that defines the digital landscape.