OpenAI to acquire Ona, extending Codex for longer tasks

OpenAI’s move to acquire Ona is a signal that AI work is shifting from quick prompts to persistent execution. The company said Ona’s secure cloud orchestration will help Codex handle longer-running tasks that stretch across hours, days, or even weeks, the kind of work that matters to engineering teams, operations staff, and enterprise buyers who need more than a chat window.

The scale behind that shift is already real. OpenAI said more than 5 million people now use Codex each week, up 400% from earlier in 2026, and that non-developers account for about 20% of users while growing more than 3 times as fast as developers. Codex, launched for macOS on February 2 and expanded to Windows on March 4, was positioned as a command center for multiple agents, parallel workflows, and long-running tasks. OpenAI also said users are increasingly running several Codex jobs at once.

For workplace teams, the important part is not just that the agent can write code or draft copy. It is that the work can continue after a laptop closes. OpenAI said remote computer use now lets Codex work in desktop apps after a Mac locks, with safeguards such as short-lived authorization, covered displays, relock on local input, and a manual-unlock fallback. That is the practical line between a demo and something a company might trust with client work, internal reporting, or production changes.

That is also why the acquisition lands so directly for monday.com. The work-OS company has already been moving toward the same operational questions. On March 11, monday.com said external AI agents could access its platform under the same governance, security, and permissions standards as human users. It also pointed to monday sidekick, monday agent builder in beta, an Agent directory, and centralized controls for access, usage limits, and asset access. On March 23, it launched Agentalent.ai, a managed marketplace for enterprises to evaluate and hire AI agents.

The overlap matters for Monday’s engineers, product managers, and sales teams because enterprise AI buyers are now asking who can actually control the machine once it starts moving. monday.com’s fiscal 2025 fourth quarter gave a clue why the company is leaning in: revenue reached $333.9 million, up 25% year over year, and customers with more than $50,000 in annual recurring revenue made up 41% of total ARR. In that market, persistence, audit trails, and approval flows are not side features. They are the product.