OpenAI’s Codex CLI just picked up what looks like a notably practical set of upgrades in v0.105, with a focus on day-to-day ergonomics: clearer output, faster input, and more control when multiple agents are involved.
What’s new in Codex CLI v0.105
The headline changes are a mix of interface polish and workflow plumbing:
- Syntax highlighting, aimed at making command-line conversations and code suggestions easier to scan as sessions get longer.
- Voice dictation for prompts by holding the spacebar, a feature that immediately got attention—and questions. Several replies suggest it’s not consistently working for everyone yet, and one commenter notes it’s labeled “still under development and is not ready for use.”
- Better multi-agent workflows, positioned as an upgrade for parallelized tasks that previously required managing multiple instances manually.
- Improved approvals controls, a change that lands squarely in the “agent safety” bucket—especially relevant for anyone uneasy about letting assistants apply changes too freely.
- “Many more quality of life changes,” per the announcement, though the post doesn’t enumerate them.
Multi-agent workflows: enthusiasm, plus the hard questions
While the update calls out multi-agent improvements, the replies also highlight what remains tricky in practice: merge conflicts between agents, competing priorities, and reports of multi-agent runs failing when processes don’t have the right privilege escalation.
Original source: https://x.com/thsottiaux/status/2027094489265807429