My AI Coding Workflow Going Into 2026

In this monologue podcast, Ray Amjad shares his AI-Agent coding workflow for 2026, emphasizing efficiency and improvement through AI tools. He details using screen recordings to generate product requirement documents (PRDs) with Gemini 3 Pro, refining them in Claude Code, and leveraging ChatGPT for identifying relevant GitHub packages. Ray discusses his role as a designer of the feedback loop, focusing on monitoring the agent's performance and updating internal documentation. He also covers model selection, preferring Opus 4.5 for large-scale features and GPT 5.2 for architecture and debugging, and shares his experiences with Claude Code and Codex CLI, as well as the use of sub-agents for research and context control. Ray also touches upon dictating prompts using HyperWhisper, utilizing Claude Code's planning mode, and forking sessions for better understanding of the agent's actions. He also promotes his Claude Code Masterclass and other tools.