Open source AI agent creator warns: fun is velocity, dependency is risk.

Peter Steinberger built an open-source AI agent that grew from a personal annoyance into 4.7 million weekly downloads in eight months. The project moved from a fun hack to a burden of security, press, and organizational complexity. He warns that feature creep and dependency on a single model became critical vulnerabilities.

When Anthropic disabled subscriptions with 24 hours notice, the project lost its optimized model and users migrated. Security reports overwhelmed him—0.3% were real threats, but the narrative damage stuck. The lesson: your dependency's business model is your business model.

Steinberger admits he stopped using his own product and lost the fun. Fun equals velocity; without it, the project stagnated into configuration bloat. He now prioritizes building for himself again, using agents proactively to review code and manage workflow.

The core insight: open source AI tools face pressure from well-funded competitors with simpler stories. The real moat is personal brand, not code. Teams should design for model independence and prepare for the governance costs that come with community adoption.