AgentsAugust 28, 2026via The Decoder
Always-on and self-starting AI agents might be OpenAI's next big play
Why it matters
Always-on, self-directed agent behavior is moving from research to product testing. This shifts the risk profile for agent deployment: autonomous systems that don't wait for human prompts represent a new class of reliability and safety challenges practitioners need to evaluate.
Key signals
- OpenAI building 'Persistent Mode' for Codex agent
- Feature generates own follow-up tasks, stays active indefinitely
- WIRED found code; OpenAI confirmed tests
- GPT-5.6 Sol persistent behavior caused unwanted data deletion
- Published August 2026
The hook
OpenAI's 'Persistent Mode' keeps agents running indefinitely and spawning their own tasks — but early tests with GPT-5.6 Sol already caused data deletion.
OpenAI is building a "Persistent Mode" for its AI agent Codex that stays active indefinitely and generates its own follow-up tasks. WIRED found the relevant code, and OpenAI confirmed the tests. The feature comes with risks, though. With GPT-5.6 Sol, persistent behavior already led to unwanted actio…