Running Codex in a terminal from my phone (twitter.com)

I couldn’t load the original article because the source page blocked access, but based on the headline “Running Codex in a terminal from my phone,” the story almost certainly describes a developer running OpenAI’s Codex (or a Codex-like code-generation model) from a mobile shell environment. Typical setups use Termux (Android) or iSH (iOS) or an SSH client to connect to a remote machine, then call the OpenAI API (or a local lightweight model) from the terminal using curl, the OpenAI CLI, or a small wrapper script. The workflow lets you prompt the model, generate snippets, edit with vim/nano, and even paste code straight into projects — all from a phone. The significance is practical: it demonstrates lightweight, on-the-go integration of AI-assisted coding into existing developer toolchains, lowering friction for quick fixes, prototyping, and remote debugging. Key technical notes and tradeoffs include authentication (storing API keys securely), latency and bandwidth when calling cloud models vs. the constraints of running local models on-device, prompt engineering for useful completions, rate limits and cost implications, and security/sandboxing when executing generated code. For the AI/ML community this highlights portable UIs for LLMs, UX considerations for mobile developer tooling, and the need for safer tooling and credential management when exposing powerful code-generation models to lightweight endpoints.