New term: agentic engineering

2026-02-04 · Publisher : Andrej Karpathy · Domain : Artificial Intelligence

Agentic engineering is professional software development in which you orchestrate and supervise agents that write the code, with no compromise on quality. Andrej Karpathy adopted it in February 2026 as the successor to vibe coding.

### Definition **Agentic engineering** is the professional practice of software development with AI agents: developers no longer write most of the code directly; they orchestrate agents, check their work and remain accountable for software quality. ### Origin On February 4, 2026, one year after launching "vibe coding", Andrej Karpathy posted a retrospective on X. He observed that programming via agents was becoming the default for professionals, "with more oversight and scrutiny", and picked a name among those proposed: "personally my current favorite "agentic engineering": "agentic" because the new default is that you are not writing the code directly 99% of the time, you are orchestrating agents who do and acting as oversight. "engineering" to emphasize that there is an art & science and expertise to it." ### What the term covers - Writing specs, reviewing plans and diffs, writing tests and evaluation loops. - Managing agent permissions and isolating them, for instance in separate worktrees. - At Sequoia AI Ascent in late April 2026, Karpathy summed up: "Vibe coding raises the floor. Agentic engineering is about extrapolating the ceiling." ### Criticism and limits Karpathy does not claim to have invented the word: he chose among several proposals. For some analysts, the term mostly formalizes a shift already under way, and the line with vibe coding remains blurry in practice. ### Key takeaway Agentic engineering marks the move from experimentation to profession. The agent writes the code, but responsibility for its quality stays human, which strengthens the case for harness engineering and specifications.

Tags : Vocabulaire, Code, Agents, Qualité