What This Is
OpenAI's latest Codex CLI update added a toggle — letting multiple AI sessions edit code simultaneously in separate working copies. It looks like a minor change; in practice, it's coding tools moving from solo operations to multi-agent parallelism.
Codex CLI is OpenAI's command-line coding assistant. When you create a new task, it now asks: continue in the current project, or open an independent "working copy" (Git's worktree mechanism) to run in.
In plain terms: you have one AI modify module A while another AI modifies module B — each works independently, without overwriting the other's changes. This isolation doesn't rely on re-cloning the codebase; it rides on Git's built-in lightweight mechanism, so multiple AI sessions can run in parallel.
Industry View
Supporters see this as the inevitable trajectory of AI coding tools toward "multi-agent collaboration" — writing code, writing tests, fixing bugs, once queued sequentially, can now run in parallel to speed things up.
But some developers are skeptical. First, this feature is essentially a wrapper around Git worktree — not technically an AI innovation. Second, having multiple AIs modify the same codebase actually increases the risk of merge conflicts, and humans still end up doing the merging. Sharper critiques go further: more code AI writes doesn't translate to faster delivery — code review costs may climb rather than fall.
Impact on the Rest of Us
For individual careers: non-programmers can sit this one out for now, but be aware that "AI collaboration" is moving from solo to team play. Other tools will catch up.
For enterprise IT: R&D teams adopting such tools will need to redesign their code review processes. AI submission frequency and parallelism will rise significantly.
For consumer markets: future tools aimed at ordinary users (not just coding) will likely grow similar multi-task isolation designs — a directional signal for product form.