# VS Code Is My AI Daily Driver. Claude Code and Codex Are My Agents.

URL: https://jeff.hopp.so/vscode-ai-daily-driver/
Category: Systems & Tools
Published: 2026-06-22
Updated: 2026-06-22

## TL;DR
The model matters, but the workbench matters more. How I use VS Code, Claude Code, Codex, Gemini, and SYNTAX as a practical AI operating environment.

## Key Takeaways
- The real shift is not choosing one model. It is putting AI inside the same workspace as the files, repos, commands, and decisions.
- VS Code has become my AI workbench because Claude Code, Codex, project context, Git, and local commands can all meet in one place.
- AI training needs to move past prompt tricks and into workbench setup, review habits, permission boundaries, and repeatable systems.

## Why use VS Code for AI work if you are not a developer?
VS Code puts files, project instructions, Git history, terminal commands, and AI agents in one place. You do not need to become a software engineer to benefit from that. You need enough fluency to work safely and review what the agent changes.

## Do you prefer Claude Code or Codex?
I use both. Claude Code is strong for broad reasoning, architecture, and confusing projects. Codex in VS Code is strong for local implementation, search, edit, command, and test loops. The workbench matters more than choosing a single favorite agent.

## Where does Gemini fit in your AI stack?
Gemini is useful as a third opinion, checker, or alternate model, but it is not usually where I start daily work. My center of gravity is VS Code with repo-native agents.

## What should businesses learn before advanced AI prompting?
They should learn how to set up a safe AI workbench: project files, source-of-truth docs, review habits, permissions, publishing rules, and repeatable workflows. Prompting matters, but the operating environment matters more.