BUILD LONG-RUNNING AI AGENTS IN SECURE CLOUD ENVIRONMENTS.

What Happened

OpenAI is acquiring Ona, a company specializing in secure, persistent cloud environments. The stated goal is to integrate these capabilities into Codex, specifically to enable the development and deployment of long-running, robust, and stateful AI agents. This move signals OpenAI's intent to move beyond ephemeral API calls to a model where agents can maintain their state, memory, and operational presence over extended periods.

Why It Matters

This acquisition tackles a fundamental hurdle for agent builders: the ephemeral nature of current agent executions. Today, agents often start from scratch with each interaction. With persistent, secure environments, agents can learn, adapt, and execute complex, multi-stage tasks over days or weeks without losing context or state. This paves the way for truly autonomous, always-on AI services that can continuously monitor, manage, and interact with complex systems, dramatically expanding use cases.

What To Build

Start designing state-aware conversational agents that remember long-term interactions and user preferences, providing truly personalized experiences. Develop agents that continuously monitor and manage infrastructure, financial portfolios, or customer service queues 24/7. Create frameworks for building persistent agent personas that evolve over time. Build tools and interfaces for managing, deploying, and observing these secure, long-running agent instances within the emerging OpenAI ecosystem.

Watch For

Keep a close eye on the concrete features, APIs, and pricing models OpenAI rolls out for these new persistent agent environments. How will they integrate Ona's security expertise to ensure robust data privacy and operational integrity? Also, monitor how competitors like Anthropic and Google respond, as persistent agent runtimes will become a key battleground.