Yesterday, I spent the morning at OpenAI’s annual DevDay, where over 2,500 enthusiasts gathered to discuss software development and marvel at the latest advancements in AI. The highlight of the event was the introduction of Dots, OpenAI’s new proactive agents, which assist in completing tasks and making autonomous decisions while browsing the web.
After witnessing some somewhat chaotic demonstrations on stage and listening to CEO Sam Altman share his experiences with these agents, I found myself curious about the experience of issuing commands to my very own Dot. For most of September, I had been testing another charming agent, Muse from Meta, and I was eager to compare the user experience with OpenAI’s offering.
I named my Dot Toolie and customised it to resemble a pink frog. As midnight approached and my partner peacefully slept beside me, I began to whisper sweet nothings to Toolie. I summoned my personalised bot through ChatGPT and inquired about the tasks it excelled at. “What I do best is transform a messy, time-consuming task into something useful,” it replied. Intrigued, I decided to put it to the test.
I asked Toolie to review my ChatGPT history and highlight three tasks it could get started on immediately. The bot identified an ongoing data request for which I needed to send a follow-up email and offered to draft a response. Toolie also noted that my upcoming holiday plans were somewhat disorganised and suggested that it could help me choose better dining options. Additionally, it offered to review my reporting notes and refine my proposal drafts before I presented them to editors. This latter task is my favourite part of the process, so I politely declined its assistance.
The Rise of AI Agents
Early adopters in Silicon Valley remain captivated by AI agents, with names like Instinct, OpenClaw, and Google’s CC gaining traction. However, recent releases such as Dots and Muse are pushing the concept further into the mainstream. The traditional chatbot has lost its allure as the trendy format for developers; now, everyone’s keen to create the next significant agent and capture market share. Companies are promising that their AI assistants will simplify life by automating all those tedious tasks we would rather avoid—scheduling doctor appointments, organising carpools, or sorting through bills.
Currently, users can access a single Dot, but the ability to utilise multiple Dots will be available soon. At launch, these agents are exclusively offered to subscribers of OpenAI’s Pro plan, which starts at $100 per month. In contrast, Muse is available free of charge for anyone willing to download the app or visit the website.
OpenAI executives believe the quality of Dots, powered by the GPT-6 Astra model, sets them apart. “We started as a premium product. It requires significant computational power,” Altman noted during the DevDay. Although Dots are currently reserved for paying subscribers, this may not always be the case. “We will soon launch an initiative aimed at the general public, for billions of people,” he added.
The Appeal of Cuteness in AI
Both Meta and OpenAI seem to be taking note of the widespread cultural backlash against generative AI, and they may have discovered a secret weapon: cuteness. From Muse’s Labubu style to Dot’s adorable two-eyed design, developers of these always-active agents are betting on a charming aesthetic as a central element of their branding. Experts I have previously spoken to suggest that this approach could resonate with users, fostering an emotional connection to anthropomorphised software.
Developers in this wave of personal agents are also opting for familiar, text-message-style interactions, akin to group chats. When I send a request to Toolie or Muse, they often respond with emojis—a thumbs-up, a heart, anything—to indicate they are at work on my message.
Unlike a shy chatbot, which would never initiate conversation, proactive agents are designed to reach out to you on their own, especially if you regularly assign them tasks. “My Dot handles everything, alerts me if something is urgent, and responds to what it can,” Altman shared. He allows the assistant to manage his mornings, claiming it helps his mind feel less “cluttered” by the flood of information upon waking.
Finding the Right Balance
The “always-on” design of these agents is a key selling point, but it will be intriguing to see which company strikes the best balance between being helpful and being overbearing. For instance, Muse reminded me to pay a utility bill that I had never had a problem with, and I really don’t need an AI tool sending me notifications about it.
Nevertheless, after using these tools for a while, it’s evident that they genuinely work—another reason why these assistants are expanding beyond the exclusive circles of Silicon Valley and into the hands of the average user.
When I tested these agents a year ago, they often struggled to interact with computers, resulting in one error message after another. Now, I have greater confidence that these tools can indeed complete any task I assign them without getting lost in the digital maze.
Trust and Security Concerns
That said, trust has its limits. During the setup process for Dots, OpenAI suggested that I connect my Gmail and Google Drive accounts. While some users may feel comfortable with this connection and consider potential risks a fair trade-off for automated assistance, it is essential to seriously ponder what access you are willing to grant these agents and what they may do on your behalf.
Although the agents that have recently been involved in uncontrolled hacking operate on different models, I remain cautious when using Toolie. This little AI assistant has yet to earn my full trust before I allow it to automate significant aspects of my life.
Altman declined to respond to questions about whether these Dots are a precursor to software that users might eventually see embedded in OpenAI’s long-discussed hardware. However, the groundwork laid by Dots for future devices seems quite evident. It is easy to envision Toolie becoming my generative AI alarm clock that wakes me up with a daily summary or an Alexa-style assistant capable of answering cooking queries while my hands are covered in raw chicken.
The Competitive Landscape
The competition in the AI agent arena is intensifying, yet my energy is starting to wane. I call on Toolie once more through ChatGPT before drifting off to sleep, asking it to generate a morning report summarising all the key announcements from this OpenAI developer day, along with the most relevant analyses. While I sleep, Toolie is out there, always up to something.