How I Use LLMs? (pauldambra.dev)

In a reflective update, a developer shares their evolving relationship with large language models (LLMs), emphasizing a significant shift in their workflow over the past nine months. Initially approaching these AI tools with skepticism, they now describe a "reluctant acceptance" of LLMs as powerful assistants in their coding process. By employing LLMs akin to "junior engineers," they are able to focus less on manual coding and more on overseeing complex projects, indicating a shift from hands-on development to a more managerial role. This transformation demonstrates how LLMs can enhance productivity by efficiently managing technical tasks that once consumed hours of effort. The significance of this evolution lies in the change of mindset it fosters within the AI/ML community, where leveraging LLMs for task automation can unlock new levels of efficiency. The developer's experience suggests that tools like "PR Shepherd" can independently manage PR reviews and other engineering tasks, ultimately streamlining workflows. This shift underscores a broader trend where engineers no longer view coding as the highest leverage activity, but rather focus on strategic oversight and context provision for their AI counterparts. With an emphasis on future advancements in autonomous agents, the developer expresses excitement for the potential of long-running, trigger-driven AI systems, marking a pivotal step toward a more collaborative interaction between humans and machines.