OpenAI AI – We put OpenAI’s new ‘Dots’ agent to the test to see if it can handle real-world errands, only to find a tool that struggles with basic recognition.
I am standing in my living room discussing couch dimensions with my partner. both of us eyeing a worn-out piece of furniture that clearly needs replacing. It felt like the perfect moment to hand the project over to OpenAI’s latest experiment: ‘Dots.’ These always-on AI agents. recently announced by the company. are designed to act as a virtual extension of the user. capable of performing online tasks like canceling reservations or booking flights.
I called the agent through ChatGPT to get the ball rolling. but the illusion of a helpful assistant evaporated the second it spoke. “Hello, Connor,” it said, greeting me by the wrong name. My partner’s skepticism, already visible, hardened instantly. It is difficult to entrust your digital life to an agent that cannot manage a basic introduction.
This technology represents a shift from niche software to something meant for ordinary people. These tools function like a chat with a friend, utilizing a virtual browser the AI controls. While Meta offers its own version. Muse. for free. OpenAI has placed its agent behind a $100-a-month subscription. though the company may eventually drop that paywall as it has with other features.
My few days with Dots revealed a product still in its infancy. The experience was defined by friction: the agent mistranscribed my instructions and repeatedly offered to solve CAPTCHAs that it ultimately couldn’t handle. These messy interactions left little room for confidence. even if the underlying concept of an autonomous assistant remains a massive. potential overhaul for how we navigate the web.
There is a peculiar tension in how these tools operate. The agents are designed to be ‘always-on,’ capable of running recurring tasks in the background and proactively messaging users. OpenAI currently limits users to one agent, though they may allow control over multiple agents in the future. To get the most out of them. the company nudges users to link personal data sources. such as a Gmail account. This move requires a significant leap of faith. as these agents are prone to privacy gaffes. and the landscape of digital automation is still largely uncharted.
When I pushed the agent—which I named Toolie—to help with the couch hunt. the process felt more like babysitting than delegating. After I corrected it on my name, I provided the specific dimensions required to fit a couch through our doorframe. Toolie then followed up by asking for our budget range. demonstrating an ability to detect that two people were participating in the chat by addressing both of us.
Despite the glitches, the intent behind these agents is clear. Companies are betting that this combination of a cuddly. welcoming persona and raw browser-control power is the next evolution of the chatbot. For now, however, the leap from a clunky, misnamed greeting to a functional digital butler remains wide.
OpenAI Dots AI agents ChatGPT technology digital assistants automation AI news