notessamuele onelia

Why Agent-Native Software Works Now

The Idea

Software agents work reliably enough today to be the substrate of real applications, not just demos. The proof is Claude Code: a large language model with access to bash and file tools, operating in a loop until an objective is achieved, can accomplish complex multi-step tasks autonomously.

The surprising discovery embedded in Claude Code’s success: a really good coding agent is actually a really good general-purpose agent. The same architecture that lets it refactor a codebase can organize your files, manage your reading list, or automate your workflows. Coding is hard enough that solving it produces an engine that generalizes.

The Claude Code SDK makes the engine accessible. You can now build applications where features are no longer code you write, but outcomes you describe, achieved by an agent with tools, operating in a loop until the outcome is reached. This opens a new category: software that works the way Claude Code works, applied to domains far beyond coding.

Why It Matters

Until now, “AI features” usually meant a chatbot bolted onto a traditional app, or a single-shot LLM call hidden behind a button. Agent-native is a different posture: the agent is the application, the UI is one of several ways to drive it, and capability emerges from composing primitives rather than enumerating features upfront.

For anyone building product right now, the question shifts from “what feature should we add?” to “what primitives let the agent accomplish anything in our domain?”

Backlinks

External sources