Updated
Updated · ZDNet · Aug 12
ZDNET's David Gewirtz Backs Both Claude Code and Codex Over a $20 Single-Winner Choice
Updated
Updated · ZDNet · Aug 12

ZDNET's David Gewirtz Backs Both Claude Code and Codex Over a $20 Single-Winner Choice

1 articles · Updated · ZDNet · Aug 12

Summary

  • David Gewirtz says Claude Code and OpenAI Codex are now close enough in capability that picking one outright matters less than matching the tool to your workflow, existing projects and preferred ecosystem.
  • Both systems can build, test, debug and modify software with long-running agents, and he argues code quality is no longer the main differentiator for most developers.
  • Pricing and bundling shape the decision: Codex comes with ChatGPT's free tier in limited form and $20 Plus plans, while Claude Code requires at least Claude's $20 tier; Gewirtz says he has used $100 Claude Max and previously spent $200 on a month of higher-tier Codex work.
  • His own setup splits by environment and project history—Claude Code mainly in the terminal for Apple app work, Codex mostly in VS Code for WordPress and security tasks—and he says using both simultaneously can be especially effective.
  • The broader takeaway is that Anthropic and OpenAI now bundle chat, coding and work agents into larger subscriptions, making ecosystem fit and hands-on trials more useful than benchmark-style comparisons.

Insights

If AI coding agents are practically identical, are tech giants just using subscription bundles to trap developers in their ecosystems?
Why do expert developers achieve double the success rate with AI coding agents if these tools are supposedly fully autonomous?
With AI autonomously executing commands, what hidden security risks lurk when developers blindly trust their terminal's invisible new co-pilot?