AI has made coding nearly free. Non-technical founders can build products. Professional developers are 10x more productive. The cost of writing code, in both learning time and execution time, is approaching zero.
What happens now?
It's important to note that this is just a continuation of a trend that has been happening for decades. Cloud computing eliminated server procurement. React and Rails compressed development time. Package managers turned complex features into npm install commands. Now, Cursor and Claude Code collapsed the gap between idea and implementation. What took weeks now takes hours.
AI might be a step function improvement but it's a continuation of falling costs, nonetheless.
Parker Conrad spent two years building Rippling before launch. He bundled payroll, benefits, HR, and IT management into one product. The integration made it better than using five separate tools. But the upfront cost was enormous.
That upfront cost just dropped by 90%. Three months instead of two years. The compound startup model, previously limited to well-funded repeat founders, is now accessible to first-time builders.
Expect a wave of bundled software across every industry. Combining three mediocre vertical SaaS products into one tight experience is suddenly viable. The integration tax disappeared.
Two visions are emerging for less complex AI-generated software. Andrej Karpathy advocates for ephemeral software: everyone builds custom apps for themselves. Mini app stores like Wabi and platforms like Lovable take the opposite approach: someone else builds, you consume from a marketplace/open web.
Ephemeral software isn't ready. Claude Code still makes mistakes. No one outside of a niche group is vibe-coding their own apps. Moreover, people like opinionated software. It considers factors you have not thought of. For ephemeral software to be big, it must be able to one-shot the perfect application, considering all the factors you have not thought of, purely from your intent. That is still a long way off.
Vibe-coded marketplaces will remain niche. The barrier isn't building cost—it's discovery. Users have 80 apps installed but use only 30 per month. Cheap AI-generated apps will flood the market, but consumers will consolidate around high-quality winners. Market dynamics won't change.
Taste. YC's adage of "build something people want" still holds true. In fact, the easier it is to build the more important it gets.