AI Coding Tools Are Flooding App Stores While User Demand Stalls

The a16z weekly chart report, as described by ifanr, found that app supply has roughly quadrupled while demand has been flat. A paper titled 'Writing Code vs. Shipping Code' by researchers at MIT and the University of Pennsylvania, published in May 2026, recorded the supply surge. From early 2025 to April 2026, the number of new iOS apps released each month roughly doubled, Chrome extensions accelerated, and Google Play new releases reversed years of decline. The researchers marked February 2025 as the start of the 'Agentic coding era.' After that point, monthly new releases rose across the three platforms: iOS went from about 40,000 to 120,000, while Android and Chrome doubled to quadrupled. The report also said that while unicorns are getting younger, the data show no payoff for independent developers' gold rush.

The demand side did not follow. The study looked at what new apps achieved in their first three months, including downloads, ratings and early users. It found that while new apps increased, total engagement in the first three months was flat or falling, and the share of apps that failed to build even a small audience was rising. On iOS, total ratings were nearly flat; on Android and Chrome, downloads were flat or declining. The share of new apps with 'zero or low use' -- fewer than 10 ratings and fewer than 100 downloads -- rose significantly, while the share showing any sign of 'escape velocity' fell. The researchers tested whether the same total demand was simply being redistributed among apps, and concluded that no new winners had emerged. Sternstein called the phenomenon App-Slop, a reference to the earlier 'AI Slop' flooding content platforms.

Sensor Tower data, cited in the same account, offered a commercial view. Using December 2024 as a base of 100, the U.S. app revenue index stood at 102.0 in August 2026, and total usage time at 107.0. In about a year and a half, revenue rose 2% and time spent rose 7%, even as more software was released. The app economy's pie did not grow with AI.

AI applications themselves are the exception. Category data showed that productivity was the only major U.S. category with both revenue and usage time growing fast. Productivity app revenue rose about 100% and time spent rose about 55%, driven by ChatGPT, Claude, Gemini and Grok. Developer tools were another small but bright category, while gaming, the largest category by volume, saw revenue decline. The article said general AI assistants benefit because different tasks -- editing email, organizing materials, discussing travel plans -- can stay in one entry point. A new app that solves only one problem must prove its value more clearly, especially if its main function can also be done inside an existing AI assistant. After ChatGPT emerged, tools such as Grammarly were notably affected. Users may pay for AI capability without being interested in every app built with AI.

Inside software development, the acceleration from AI has not fully reached delivery. The study combined GitHub activity and AI tool usage data from more than 100,000 developers. Researchers estimated that cumulative adoption of different generations of AI coding tools increased code commits by about 180% and the number of software releases by about 30%. The article described the remaining work: a developer can build a bookkeeping app with income, expenses, categories and monthly charts, but must still decide whom it serves. A college student trying to control takeout spending and a family managing shared expenses need different features. Commercialization brings more questions: how will potential users learn about the app, how much does it cost to acquire a paying user, and can subscription revenue cover model calls plus service costs?

Even among people who pay for AI, the consumer market remains small. Consumer Edge data cited by a16z showed that, as of the first quarter of 2026, about 3% of U.S. consumers personally paid for AI services. The share was about 5.9% among people aged 18 to 24, about 5% among those 25 to 34, and only 1.4% among those 65 and older. Young people were paying for AI at roughly four times the rate of older groups. Over the past year, the share of cardholders paying for at least one AI service roughly doubled, but direct paying users remained limited.

The article also reviewed the startup landscape. New unicorns are getting younger, with a median age of four years, while older companies are getting older, and the middle layer is disappearing. Citing SVB data, a16z said median revenue growth across several technology sectors it tracks fell from about 40% to 70% since early 2022 to about 15% to 30%, while losses narrowed. Companies slowed expansion and focused more on improving profit and extending cash runway. Companies that recently raised still had higher growth and larger losses. SVB estimated, based on first-half progress, that about 2,345 venture-backed U.S. companies could fail or close in 2026. More than one-third of the failures it tracked in 2026 were founded during the 2019-2021 zero-interest-rate period.

When anyone can make an app in hours or even minutes, building the thing itself becomes a harder business. AI coding has changed the software supply curve: small needs that once went unserved because development was too expensive can now be turned into products quickly. App stores will become more crowded, and software may be generated in bulk like web pages, images and video. But users still have only 24 hours a day. They will not download four times as many apps because supply has quadrupled, nor pay four times as much because developers finished in one sentence what once took a month. The article argued that the value of vibe coding is in lowering the cost of validating an idea, not the cost of turning it into a business. A minimum viable product can be tested with real users, and data can show whether demand exists before more investment. The same logic applies to tokenmaxxing in the large-model industry: more tokens and more calls do not automatically become more revenue, and more code and more apps do not automatically create more demand. For independent developers, the situation may be harsher. Many have become users of AI applications, and former paying users have become independent developers, but the apps those developers build with AI have not found new paying users. AI lowers the barrier to entry and also lowers the value of many simple software products.