Karpathy Declares 'Vibe Coding' Obsolete, Pushes Agentic Engineering After Joining Anthropic in May 2026
Updated
Updated · Forbes · Jun 12
Karpathy Declares 'Vibe Coding' Obsolete, Pushes Agentic Engineering After Joining Anthropic in May 2026
3 articles · Updated · Forbes · Jun 12
Summary
February 2026 marked Andrej Karpathy’s break with “vibe coding,” a term he coined a year earlier, as he recast AI software work as “agentic engineering” and then joined Anthropic in May to rebuild its pretraining team.
The shift reflects how coding tools have moved beyond plain-English prototyping toward agents that can execute larger chunks of work with limited human intervention—letting users set goals, review outputs and iterate instead of writing most code directly.
Bloomberg reported senior engineers at three companies were already working this way, while Harvard researchers used agents to build internal tools that non-coders could brief and receive within about 1 hour.
For startups, Karpathy argues the payoff is speed and scope: a feature can shrink from 1 week to 1 day, and PYMNTS said a U.S. neobank shipped an agent-built customer feature after a 4-hour compliance review instead of a 4-month build.
That makes model quality the bottleneck, strengthening Anthropic’s strategic position as companies compete on which underlying model lets agents finish the most work on the first pass.
As AI agents proliferate, are we building a productivity revolution or the world's largest, most unsecured attack surface?
If an AI agent commits a crime, who is held accountable: its user, its creator, or the machine itself?
From Vibe Coding to Agentic Engineering: The 2026 AI Shift Driving 4% of GitHub Commits and Reshaping Developer Roles
Overview
In May 2026, Andrej Karpathy, a co-founder of OpenAI known for his work in deep learning and computer vision, made a strategic move to Anthropic. He expressed strong enthusiasm for joining the team, highlighting his excitement to return to hands-on research and development. Karpathy believes the coming years will be formative for large language models, signaling a focus on cutting-edge advancements. While he remains passionate about education and plans to return to it later, his immediate impact at Anthropic is expected to drive innovation in AI, intensifying competition with OpenAI and shaping the future of agentic engineering.