OpenAI's Agent Ambitions vs. Reality

OpenAI is currently leaning into the hype, with president Greg Brockman recently suggesting it is not unreasonable to feel the company has entered the "AGI era" via the GPT-6 Astra model. But looking past the marketing, as first reported by The Verge, OpenAI is lagging in the race for continuously running, consumer-facing AI agents—a category where competitors are already delivering on long-standing industry promises.

According to reporting from The Verge, rivals like Meta’s Muse have already captured significant traction, topping App Store charts with 600,000 daily active users in the U.S. per Apptopia data. Other contenders include Google’s Gemini Spark, which boasts over 30 external service partners including Spotify, Uber, and Dropbox, and the open-source OpenClaw. These systems are reportedly handling complex real-world workflows, from canceling subscriptions and paying bills to organizing group trips and DMV appointments.

Rumors suggest OpenAI will announce its own agent, dubbed Aeon, at its upcoming DevDay. While OpenAI has employed Peter Steinberger, the creator of OpenClaw, the company faces a steep climb. In my view, a spec sheet won't save them here; Aeon needs to match the ease of use seen in Instinct or the integration depth of Meta and Google to be viable.

Furthermore, the path to success is fraught with the same security risks that have plagued the category. The Verge notes that Muse has faced complaints regarding the unauthorized sharing of home addresses and private messages, while a now-patched vulnerability previously allowed attackers to seize control of accounts. If Aeon is to succeed, it cannot just be another glorified chatbot; it must solve these autonomy risks while competing under the pressure of a looming IPO.