Meet Atlas
OpenAI made good on months of rumors that they would launch their own browser. At the core, the browser is familiar enough to Chrome and Safari that the average user would find it welcoming. Under the hood, it is a step function ahead of Chrome when it comes to memory and agents.
Memory: The idea is that Atlas can look over and remember everything inside your browser. That includes web search activity, Gmail, Docs, and hosted apps. That’s important because, in theory, it should deliver a highly personalized experience, specifically the ability to find what you’re looking for, whether it’s an old web page, hosted doc, or email. I believe there is significant value in a browser with memory, and that Atlas will push Google to get more aggressive at building memory and personalization in Chrome.
Agents: The agent mode within Atlas felt a lot like the Operator agent that OpenAI launched last January. The two demos, converting a doc into a CRM tool and planning a shopping list. were similar to use cases we’ve played with in Operator earlier this year. Over time, agents will unlock real AI utility, giving the ability to direct your browser to do work online. Importantly, it only accesses info on your browser, hosted apps, and the web. Google is well aware of the role agents will play, and I expect they will quickly respond with similar features.
While there was no mention of monetization, the topic of OpenAI’s impact on Google’s search business still looms large.
