A detailed personal workflow breakdown of how to use OpenAI Codex beyond simple coding tasks. The author describes building long-running 'megathreads' with persistent memory stored in an Obsidian vault (synced as a GitHub repo), using voice input for richer context, steering agents mid-task, and setting up Heartbeats for recurring automated checks across Slack, Gmail, and other tools. Key patterns include using Goals with verifiable success criteria (e.g., migrating Python Rich to Rust with passing unit tests), leveraging the side panel for artifact inspection and annotation, and combining connectors ($slack, $gmail, @chrome, @computer) to create cross-tool feedback loops that keep working after the user steps away.
Table of contents
Durable threads ¶Voice input ¶Steering ¶Memory ¶Computer and Browser Use ¶Remote control ¶Heartbeats ¶Goals ¶The side panel ¶Comments216 Impressions