What comes after the code? Event recap: Northflank x Augment Code x Zed

A panel discussion in San Francisco, involving engineers and founders from companies like Augment Code, Zed, and Human Layer, explored the current capabilities and limitations of coding agents in software development. The conversation, moderated by Steve Jang and co-hosted by Kindred Ventures, emphasized that while coding agents can handle tasks such as typing code, the critical decision-making and design aspects remain with human engineers. The panelists highlighted the importance of well-defined, scoped tasks and the need for tools like linters and test suites to ensure the quality of agent-generated code. The discussion further touched on the structural cost advantages of subsidized services like Claude Code and the challenges of integrating large language models (LLMs) into production environments. Northflank's CEO, Will, described how their platform automates aspects of the software lifecycle beyond code generation, focusing on deployment and infrastructure management to handle the rapid pace of agent-generated pull requests. The economic and compliance considerations of self-hosting models were also discussed, with a view that LLMs should be integrated into the broader infrastructure stack. The panel concluded with insights into Northflank's focus on sandbox environments and preview setups to safely manage and verify agent-generated code before production deployment.