Why IDEs Won’t Die: From Editing Files to Managing AI Agents

Andrej Karpathy, a leading AI‑coding advocate, says IDEs will not disappear but must change their usage to accommodate AI agents.

Traditionally, an IDE’s job is to edit, save, and compile individual source files such as *.py or *.js, with developers manually linking modules to build a runnable system.

In the AI era, code can be generated automatically; the fundamental programming unit becomes an autonomous agent that executes tasks when given clear goals and constraints, eliminating the need to write low‑level code line by line.

Consequently, the next‑generation IDE should be an Agent Command Center that manages, schedules, and monitors many AI agents, rather than merely handling files.

Key capabilities of this larger IDE include:

One‑click show/hide of any agent to keep the interface tidy.

Real‑time status display, indicating idle, active, or error states for each agent.

Instant tool pop‑ups, such as opening a terminal for a specific agent.

Detailed usage statistics: token consumption, compute resources, runtime, and task completion rates.

Full‑screen or multi‑monitor layouts that present a comprehensive agent overview.

Karpathy illustrates the concept with a humorous diagram of corporate org charts, calling them “org code,” and suggests future IDEs will be used to construct, run, and manage such organizational structures.

The community has responded enthusiastically, noting that the new paradigm justifies larger displays and a shift in how developers interact with their tools.