The latest a16z weekly data report has poured cold water on the “vibe coding startup wave”: app supply has quadrupled, yet demand has barely moved. A paper published in May by researchers at MIT and the University of Pennsylvania, titled “Writing Code vs. Shipping Code,” shows that from early 2025 to April 2026, the number of new iOS apps released each month roughly doubled.
The researchers mark February 2025 as the starting point of the “agentic coding era.” After that, monthly new app launches on all three major platforms took off: iOS jumped from about 40,000 per month to 120,000, while Android and Chrome each doubled to quadrupled. The curve became nearly vertical after the inflection point.
But looking at the first three months of these new apps, the results are grim: the number of new apps has risen significantly, yet total engagement captured by newly released apps each month has remained flat or declined. The share of apps that fail even to build a small audience is still rising. The researchers named this phenomenon App-Slop.
Structural data makes the problem clearer. When segmented by usage, the share of new apps with “zero or low usage”—fewer than 10 ratings and under 100 downloads—has risen markedly, while the share of apps showing any “escape velocity” and crossing the 10/100 threshold has fallen sharply. The researchers specifically tested whether the total pie was unchanged and only market share was being redistributed. The answer was no: it is not that the pie stayed the same, but that no new winners emerged at all.
This shows that AI has lowered only the production barrier; delivery and deployment still require judgment. StarWar Cloud's focus in enterprise AI deployment and AI training is precisely on how, after mastering tools, to turn technology into genuinely usable products and business outcomes, rather than stopping at a prototype generated by a single prompt.
Interestingly, AI itself is thriving. By category, productivity is the only U.S. market category seeing both revenue and time spent grow rapidly, driven by ChatGPT, Claude, Gemini, and Grok. New apps that solve only a single problem need to prove their value more clearly. Combining data from more than 100,000 developers, the study also estimates that after adopting AI coding tools, code commit activity increased by about 180%, while the number of software releases increased by only about 30%.
From “can it be built?” to “will anyone use it?” lies a large amount of work: whom to serve, how to acquire customers