Vibe Coding
Codex CLI 0.153.0 adds a hidden context-management mode with a `new_context` tool that skips compaction entirely
Released 2026-09-03, `features.context_management.experimental_mode` is disabled by default and, when enabled for eligible ChatGPT Plus/Pro/Pro Lite sessions on the Codex backend, activates token-budget context, history notes, and a `new_context` tool that lets the model request a fresh context window without spending tokens on a compaction summary. API-key sessions, custom providers and temporary structured threads are excluded. This is a different bet from every compaction-based harness: throw the window away rather than summarize it.
Source
↳ Follow the thread