Deep Dive
The core win: tasks that survive your laptop shutdown
The demo opens with a casual admission — the presenter is building something in Codex and needs to close the window. Instead of killing the process, he kicks the task to the cloud. Once it's running remotely, he closes his terminal and the work keeps going. This is the opposite of local development, where closing your IDE stops everything. The freedom here is immediate: you can walk away, use your phone to check on progress, even voice-control new instructions into the same task without interrupting it. It's the difference between being tethered to a terminal and being able to actually move around.
Smart onboarding that reads your code
When you create a new cloud environment, a chat interface guides you through setup. Instead of you manually listing dependencies or guessing at networking requirements, the system reads your repository and researches what it needs. This covers everything from Node modules to container networking to secrets management. Users can share environments with their team, and teammates can override secrets with their own vault credentials instead of baking API keys into the config. Once validated, you publish the environment as a named snapshot. Now your whole team can spawn new tasks using that exact configuration without redoing any setup work.
Delegating code review to Dot while you do other things
After the cloud task completes, the presenter demos the new code review flow built into Codex. Instead of manually running tests and reviewing the PR himself, he tells Dot (an AI assistant) to handle it — create a PR, run tests, flag anything that needs approval. Dot gets to work automatically while the presenter leaves to buy a laser machine. This ties the cloud environment feature to a broader productivity loop: you can stop context-switching between your local work and async review because the AI handles the gatekeeping.