# What AI-Native Full-Stack Means

**Track:** Foundations of AI-Native Development — AI-Native Full-Stack Development — complete (30)
**Framework / surface:** web dev (web architecture)
**Level:** Beginner
**Prerequisites:** none

**In one line:** AI is not an add-on: it changes how software is built, what runs inside it, and who — or what — can use it.

## Theory, aesthetics & inspiration

The claim is broader than "use AI to help write code." AI now enters the web stack in three places at once: a coding agent joins the development process — reading the repository, editing many files, running commands, browsing the application it just changed; models and agents become runtime components alongside the database and the queue; and the product grows a second, machine-readable face, because its next user may be another agent. Andrej Karpathy's "Software 2.0" essay (2017) anticipated the turn — software increasingly specified by data and optimization rather than hand-written logic — and by 2026 the frameworks had caught up, with Next.js publicly rebuilding its developer experience "for an agentic future." The durable skill is therefore architectural literacy — boundaries, state, permissions, failure modes, verification — directed at both deterministic code and probabilistic collaborators. The reference stack will change; the conceptual stack outlives every implementation of it.

**Builder question:** Which parts of your application should remain deterministic, and which genuinely benefit from model judgment?