OpenAI shows Codex's new workflow: local-cloud handoff, voice control, dots

The workshop came as coding agents from OpenAI, Anthropic and Google DeepMind have been getting constant product upgrades, but the shift to cloud has not gone smoothly for users of Codex, Claude and Antigravity. Codex itself launched in May 2025 as a cloud-based agent that can work on many tasks in parallel, yet Kundel admitted it took a while for models and harnesses to reach a really engaging experience. That began to change after DevDay in late-September, when a revamped version 2 of Codex Cloud was announced with the promise that you can finally close your laptop and your agents will keep working. The new cloud environments let a team publish a setup once and run multiple tasks in isolated instances, selecting the GitHub repositories to access.

Within that evolution, the workshop positioned voice and dots as complementary interfaces rather than replacements. Guo described voice as a control plane for managing threads and pulling data from other apps, while Kundel said he often talks to his dot on the way to work to start pull requests before reviewing them in the office. Guo also acknowledged that dots feel awkward to Codex power users because they are a new paradigm good at different things. Kundel closed with tips including turning down the effort level and asking agents for help, but his main point was not to get too attached to any current way of working, since things are constantly changing.