Vibe Coding: Maintaining Productivity Without Typing
Theo - t3.gggo watch the original →
the gist
A hand injury forced a shift from high-speed typing to 'vibe coding'—using voice-to-text, AI agents, and remote orchestration to manage computer tasks without manual input.
The Shift to Voice-First Engineering
After a hand injury limited his ability to type, Theo transitioned from traditional keyboard-heavy development to a voice-and-agent-driven workflow. He emphasizes that 'coding without typing' is a misnomer; the real goal is 'vibe coding'—using AI to handle the mechanical navigation and execution of tasks that would otherwise require manual input. By leveraging voice-to-text tools like Whisper Flow and delegating complex multi-step processes to AI agents, he maintained and even increased his shipping velocity.
Optimizing the Voice-to-Text Environment
For voice-to-text to be viable in an office setting, the hardware must be optimized for privacy and background noise rejection. Theo discovered that a dedicated, inexpensive podium microphone (approx. $70) allows for whispering commands at a volume that is inaudible to nearby colleagues while remaining crystal clear to the AI. This setup solves the social friction of 'yapping' at a computer in a shared space and provides superior noise isolation compared to standard laptop microphones.
Remote Orchestration and Agentic Workflows
To avoid the 'hell' of terminal interaction without a keyboard, Theo relies on agentic orchestration tools like 'Fleet' and 'T3 Code'. These tools allow him to manage multiple machines and complex file operations (e.g., transferring large files or building custom software branches) by providing high-level natural language instructions. By pulling AI into the workflow earlier—instructing agents to find, download, build, and organize files across different environments—he reduces the need for manual context switching and terminal commands.
Separating Domains for AI Assistance
Theo maintains a strict separation between his coding and life-management workflows. He uses T3 Code for all development-related tasks and ChatGPT for personal administrative tasks, such as managing medical records. This separation prevents his development environment from becoming cluttered with non-code context while ensuring that AI agents have the appropriate scope to handle tedious, repetitive tasks like downloading and uploading dozens of medical PDFs.