Mitchell Hashimoto outlines his journey with AI adoption, emphasizing three distinct phases: inefficiency, adequacy, and workflow discovery. Initially, he struggled with the transition, finding it difficult to abandon familiar workflows for new tools. His first significant AI experience was with a chatbot, which he found limited in coding tasks. While impressed by its ability to recreate a command palette, he faced frustration when attempting to use it for other coding projects, highlighting the inefficiency of relying solely on chat interfaces for meaningful work. Hashimoto then transitioned to using AI agents, which are more capable due to their ability to read files, execute commands, and perform HTTP requests. He had to push himself to reproduce his manual coding tasks with the AI to understand its capabilities fully. This process was tedious, but he gained insights into optimizing interactions with agents. He learned to break down tasks, separate planning from execution, and allow agents to verify their outputs. These adjustments led to a more effective integration of AI into his workflow, though he initially didn’t notice significant time savings. To further enhance efficiency, he dedicated the last 30 minutes of each workday to set up agents for various tasks, such as deep research and issue triage. Although he initially found this practice unproductive, he discovered that it allowed for substantial progress during downtime. By the end of his journey, Hashimoto became adept at identifying tasks suitable for AI, allowing him to focus on more complex work while agents handled routine tasks. He emphasized the importance of managing notifications to maintain deep focus during work.