Foundation Capital introduced the concept of Context Graphs, describing them as dynamic records of decision traces across time and entities. These graphs evolve into a reliable source of truth, explaining not just what occurred, but why it happened. In the realm of software development, the shift from bandwidth constraints to context constraints has changed how developers collaborate. Git, created in 2005, allowed minimal commits due to the limitations of email-based code sharing. Now, with the abundance of data, the challenge lies in managing context effectively. Agent Trace addresses this issue by linking code changes directly to the conversations and contexts that inspired them. Developers can associate each change, whether a Git commit or other adjustments, with specific discussions, making it easier to retrieve relevant background information. This system not only helps organize code but also protects sensitive information by filtering out personal data from the agent trace store. The article highlights various tools stemming from Agent Trace, such as a file viewer that distinguishes between AI and human contributions and new interfaces for pull request reviews, all aimed at improving visibility in development processes. The potential for Agent Traces is significant, especially in environments with multiple contributors, including humans and AI. By implementing these traces, organizations can utilize management dashboards to make informed decisions based on comprehensive data. The article emphasizes that the value of development will shift from merely counting lines of code to prioritizing context, which will enhance efficiency and effectiveness in programming. In the coming years, improvements in AI models like GPT-5 will underscore the importance of hidden reasoning artifacts, further validating the need for context in development. This approach aims to enhance the capabilities of coding agents, ultimately leading to smarter, more efficient software development practices.