Google says Gemini users will not have to know whether their task requires Spark, Daily Brief, or simply a quick search through email. But according to Sarah Perez’s opinion in TechCrunch, this promise reveals a contradiction within the app itself: nearly every capability gets its own name, identity, icon, and separate space, making users learn the product’s structure instead of focusing on the task they want to accomplish.
In the Gemini app, users can switch between Chat, Spark, and Daily Brief, which appear as separate functions within the interface. Perez argues that this division adds clutter to an experience that could have been simpler, and suggests that Gemini has not yet settled on the central function by which it wants to be known to ordinary users.
When Proactive Assistance Becomes Annoying
Daily Brief offers an example of the gap between what seems intelligent from an engineering perspective and what appears useful in everyday use. The function creates something resembling an AI-powered agenda and presents “proactive and personalized” updates based on data from Google apps such as Gmail and Calendar.
But the article indicates that Daily Brief does not always distinguish between urgent or actionable information and reminders the user did not request. It may repeat a suggestion to follow up on research started in the chatbot, or display previous searches on Google. Perez describes this second type as potentially unsettling, because searching for a scholarship or for animal rescue does not necessarily mean the user wants the service to pursue those topics with them later.
A Practical Feature Hidden Behind a Brand
On the other hand, the writer considers Spark one of the most useful parts of Gemini, because it functions as an AI agent capable of taking actions on the user’s behalf. However, Google presented it as a separate brand, forcing users to think about where they should go before writing their request.
The editorial argument here is clear: users should not have to decide in advance which “section” within the app is appropriate for the task. They should be able to write or say what they want, after which the system determines internally whether it needs an ordinary conversation or an agent capable of carrying out additional steps. These labels may be useful to engineering teams or for internal organization, but they are not necessarily useful for an interface aimed at a broad audience.
The Problem Goes Beyond Gemini
Perez does not limit the problem to Google. She points out that the AI industry exposes its internal structure to users instead of hiding it behind a simpler experience. In Anthropic’s Claude app, users must choose between “Chat” and “Cowork,” while ChatGPT requires navigation between “Chat” and “Work.” The writer describes this as a design based on an engineering way of thinking, asking consumers to learn names for different interfaces or interaction modes even though they operate on the company’s underlying AI model.
The article notes that the problem is not limited to names, but also involves the fragmentation of context. As of that week, the two Claude modes did not share previous conversation memory within the app, according to the text. This makes the separation between functions more apparent from the user’s perspective, and perhaps more confusing.
Why Could Siri’s Simplicity Be an Advantage?
Perez believes Apple’s less showy approach to Siri may ultimately win consumers’ trust, because owners of iPhones and Apple devices do not need to change their behavior or learn a new interface. Instead, the company adds smarter capabilities to tools they already use, such as Spotlight Search, the Photos app, the iPhone camera, and voice requests to Siri.
The same principle applies, according to the article’s interpretation, to the rise of text-message-based AI services. Services such as Poke, Ollie, Lindy, Orchid, Lucas, Folk, Tomo, and Instinct allow users to send a message to an assistant and let it handle the request, without searching for the right feature inside a larger app.
A text-message interface is familiar and does not require users to memorize the product map or understand the differences between agents and conversation modes. The article quotes Justine Moore, an investment partner at a16z, as saying that people do not want to open an app every time they need help; they want a contact they can message like a friend, and that iMessage represents the highest standard for this type of interaction.
Editorial reading: What is really changing is not only the models’ capabilities, but also how they are presented. As functions multiply, hiding internal complexity becomes part of product quality, not merely a cosmetic interface decision. But the source does not prove that one model or one particular approach will necessarily prevail; proactive interaction also remains tied to a sensitive question about the limits of using email, calendar, and search-history data. The issue therefore appears open between the convenience of an assistant that acts automatically and the risk that it will become a source of alerts or suggestions the user did not request.