What ClickHouse learned from a year of coding with AI agents (thenewstack.io)

ClickHouse has shared insights from a year of utilizing AI agents for coding, revealing a significant evolution in their effectiveness and application in software development. The company categorized AI-assisted coding into three levels: basic chat interactions, integration into development environments (CLI or IDE), and fully autonomous agents in isolated settings. While initial skepticism surrounded the use of agents for complex tasks in ClickHouse’s C++ codebase, the release of Claude Opus 4.5 marked a turning point. Agents demonstrated their capability in handling boilerplate code, resolving merge conflicts, conducting code reviews, and investigating bugs—activities where they outperformed human programmers. This growing reliance on AI agents represents a shift in the AI/ML community, highlighting how well-implemented agents can amplify productivity and code quality. ClickHouse’s experience emphasizes treating AI as an augmentative tool rather than a replacement for human judgment. They advise starting with simpler tasks and gradually increasing complexity while validating agent outputs to mitigate risks. The journey reflects the maturation of AI tools in real-world applications, suggesting that developers and teams embracing these advancements stand to benefit considerably, making it a pivotal moment in AI-assisted software development.