Agentic coding tools rewrote the supply side of software
Agentic coding tools are AI systems that take high-level instructions, write and debug code, and ship working applications with limited human input, which has sharply increased software output while exposing that real value lies in user adoption and durable workflows rather than raw code volume. SimilarWeb data shows the impact on mobile app stores: iOS and Android launches jumped about 50% year-over-year once agentic coding matured in early 2025, reversing several years of shrinking release growth. From mid‑2022 to early 2025, app release growth had fallen as low as around -35% year-over-year, then spiked to nearly 55% growth by early 2026 as AI-generated apps poured in. The barrier to releasing software has collapsed; a small team can now ship an app in days. But the numbers also reveal that lower friction to ship does not guarantee there is a market that wants what gets built.
A flood of AI-generated apps meets a wall of user indifference
Despite the boom in developer productivity, adoption has not kept pace. SimilarWeb’s download data for Android apps released since February 2025 shows that 75.2% have failed to reach 1,000 total downloads, while only 2.7% have crossed 100,000 downloads. Most AI-generated apps land in near‑invisible territory, buried in app store lists with minimal engagement. This pattern is not entirely new—most apps have always struggled—but the volume is: agentic coding tools are amplifying an existing discovery problem into app market saturation. As George Hotz warned, AI coding agents risk creating “a golden era for buckets and buckets of slop,” where rapid shipping produces many more experiments than enduring products. In this environment, the scarce resource is not code or features but sustained user attention, trust and habit, which are much harder to automate.
Why more code doesn’t equal better products
The disconnect between output and usage exposes a quality-over-quantity issue at the heart of AI-generated software. Agentic coding tools excel at turning a vague idea into shippable code, but they do not validate whether that idea solves a pressing problem or fits into a real workflow. Investors are already recalibrating their expectations. As Ivan Nikkhoo notes, growth-at-all-costs has given way to scrutiny of customer retention, capital efficiency and measurable ROI. A flashy demo is no longer enough; buyers want proof that a product owns a critical workflow, not a single use case. This is especially true when switching costs are low and competitors can replicate features in months. Without a clear buyer, a sharp wedge into a painful job, and strong usage patterns, AI-coded apps risk becoming disposable experiments rather than compounding assets.
From features to defensible workflows and business outcomes
For founders using agentic coding tools, the playbook must shift from shipping speed to defensible workflow ownership. Instead of building one more generic assistant, teams need to design systems of intelligence or vertical operating systems that sit in the center of a customer’s day-to-day operations. In the new SaaS reality described by Nikkhoo, buyers are less interested in seats and more in outcomes: code written, tickets resolved, contracts reviewed or back‑office workflows automated. Pricing is following that logic, moving toward usage-, consumption- and outcome-based models because AI can perform work directly. That means the winning AI-generated apps will be those that tie usage tightly to a clear business result, demonstrate retention under budget pressure, and expand from a narrow wedge into a broader platform. Developer productivity is now table stakes; workflow depth and economic impact are the moat.
Cloud choices and BYOC become the hidden moat for AI apps
As AI-generated apps multiply, cloud infrastructure strategy is emerging as a less visible but important differentiator. When anyone can spin up an agentic coding workflow, long-term advantage comes from how reliably, securely and cheaply those agents run at scale. BYOC (bring your own cloud) models, where customers deploy AI systems into their own cloud environments, are likely to gain weight for enterprise buyers that worry about data control and lock‑in. This aligns with investor focus on long-term margins and defensibility: cost-of-intelligence improvements and flexible deployment can support better unit economics over time. For developers, the lesson is that shipping an app is only step one. To stand out in an era of app market saturation, AI-generated apps must combine agentic coding tools with thoughtful cloud architecture, strong data ownership stories and business models that tie infrastructure spend to clear outcomes.
