Meta is preparing to launch Hatch, a consumer AI agent platform that can interact with outside services and execute real tasks, while a future model codenamed Watermelon aims to upgrade its reasoning core.
The push marks a shift in Meta’s AI strategy. Rather than treating assistants as conversational add-ons inside Facebook and Instagram, the company wants them to operate across services such as DoorDash, Etsy, Outlook, and Reddit. That is a harder engineering challenge, because it requires reliable API interactions, account security, and clearer accountability when an agent makes a bad decision on a user’s behalf.
Hatch also arrives amid broader pressure to show AI returns beyond advertising. Meta has invested heavily in data-center expansion and model development, and investors are watching whether the company can translate that spending into subscription revenue or higher engagement. Early reports of premium tiers near $199.99 per month suggest Meta is testing whether consumers will pay for an agent that does real work instead of merely generating text.
Watermelon, the expected follow-up model, is still months away, but its anticipated October launch gives Meta a roadmap: first ship an agent wrapper, then improve the underlying intelligence. If Hatch gains traction, it could become a meaningful test of whether consumers trust AI with transactional decisions. If it stumbles, it will reinforce skepticism that Big Tech is rushing to productize agents before the technology is ready.
Image source: i.ibb.co