OpenAI built the chatbot era, but it’s now playing catch-up in the fastest-moving part of AI: always-on, consumer-facing agents. According to The Verge AI, OpenAI is widely rumored to launch its own agent, reportedly called Aeon, at its 2026 DevDay on Tuesday. It’ll step into a crowded field that already includes Meta’s Muse, Google’s Gemini Spark, SpaceX’s Grok Bot, Instinct and the open-source OpenClaw.
What stands out is how fast the ground has moved. A year ago, OpenAI looked like it would define this category. Today it has to prove it still belongs in the top tier.
🧭 How OpenAI Lost the Lead
OpenAI moved early. It demoed an agentic API at DevDay 2024 and shipped early agents for computer use and research in mid-2025. But The Verge AI notes that agents only went “from proof of concept to practical success” late last year, when OpenClaw arrived. Since then, competitors have raced ahead while OpenAI’s visible progress has stalled.
Meta is now the main rival. Muse launched earlier this month, topped the App Store charts and reached 600,000 daily active users in the US, according to Apptopia data cited by The Verge AI. Reviewers call it useful but “slightly creepy.” Instinct has won fans because it’s easy to use, and you can talk to it over iMessage and other chat apps. Google’s Gemini Spark runs 24/7 and connects to more than 30 partners, including Dropbox, Uber and Spotify.
These agents aren’t demos anymore. Users report them filling out doctor’s forms, canceling subscriptions, booking DMV appointments and paying bills.
🧩 What a Winning Agent Needs
The Verge AI argues that Aeon has to combine the best parts of several products to win:
- Capability: as useful as OpenClaw
- Price and setup: as cheap and easy as Muse
- Interface: as easy to chat with as Instinct
- Security: far safer than anything shipping today
That last point may matter most. At least one now-patched vulnerability let attackers take over accounts, and some users say Muse gave out a home address without permission and read private messages. Those stories are hard to verify, but they point to a real problem. The more you let an agent do on its own, the more damage it can cause when something goes wrong.
⚖️ OpenAI’s Cards: Strong Model, Weak Ecosystem
OpenAI has two clear advantages. First, it employs Peter Steinberger, the creator of OpenClaw, who’s very likely helping build Aeon. Second, Aeon would probably run on GPT-6 Astra, the frontier model OpenAI has been hyping hard. “I think it’s not unreasonable to feel that we are now in the AGI era,” OpenAI president Greg Brockman said at a press briefing this month.
Treat that claim with some caution. OpenAI has a long history of big launch-week promises, and “AGI” has become a marketing word as much as a technical one. A stronger model should help with coding, research and computer use. It won’t fix the distribution problem, though.
That’s where OpenAI is weakest. Muse plugs into Meta apps used by billions of people. Gemini Spark sits on top of Google’s huge set of personal and enterprise services. OpenAI’s answer so far has been the ChatGPT Apps SDK, which brought Zillow, Spotify and Canva into ChatGPT last year. That’s a start, but it doesn’t match an ecosystem people already live in every day.
🔮 The Next 1-3 Years
I expect the agent market to split along three lines:
- Ecosystem owners (Meta, Google) win on convenience and cost because the integrations already exist.
- Model leaders (OpenAI) win on hard tasks, where reasoning quality matters more than reach.
- Trust becomes the tiebreaker. The first major agent scandal involving money or personal data could reshape the rankings overnight and invite regulators in.
OpenAI also has its upcoming IPO to think about. An agent that ties together Astra, the Jony Ive hardware project and ChatGPT’s user base would make a strong story for investors. A weak launch would raise tough questions.
🛠️ What You Should Do Now
- Businesses: Make your services agent-friendly. Clean APIs and clear booking flows will decide whether agents send customers your way or to a competitor.
- Practitioners: Give agents the least access they need. Watch what they read and share, and require human approval for payments and anything involving personal data.
- Buyers: Don’t commit to one platform yet. This market could look completely different within six months.
Tuesday will show whether OpenAI can turn its model lead into a product people actually trust. The Verge AI has the full rundown of the rumors ahead of DevDay.