Google continues to expand its Gemini platform with new options for custom gems, giving users more granular control over how these agents operate. In the latest build, the most notable addition is a tool selector in the Gem editor, allowing creators to pick which capability, like image or video generation, a Gem will default to when users start a conversation. This replicates the menu from standard Gemini chat, but now custom Gems can leverage these tools from the outset.
The real story, though, is the expanded list of tools visible in this menu, revealing features that are not yet public. One unreleased tool, simply called “Agent,” suggests Google is moving towards deeper web navigation and research workflows, comparable to how ChatGPT’s advanced models can browse and interact with web content. Evidence of this Agent tool has appeared in recent code updates, flagged as a labs feature and likely undergoing internal trials. If implemented, this would broaden Gemini’s scope beyond static generation tasks into more dynamic information gathering and automation.
Another upcoming feature is “Canvas Creative,” presumably tied to the Canvas experiment in Labs, where users can place generated images on an interactive canvas. Early UI elements, such as animated expanding boxes and hints at CSS-driven layouts, imply this could evolve into a more interactive or even collaborative workspace for visuals.
“Visual Layout” is also mentioned, though its purpose is less clear. It could be an extension of the Canvas feature, potentially allowing for richer or multi-modal layouts within the Gemini interface.
Finally, there’s the “Deep Think IMO” option, which may allow Gems to perform more sophisticated reasoning or analysis. While it’s uncertain whether this will reach custom Gems, its presence hints at more modular and targeted workflows for power users.
For now, there’s no concrete release timeline. These developments reflect Google’s ongoing strategy to close feature gaps with competing platforms by making Gemini a more customizable, agent-driven environment. If these tools reach public release, they’ll likely appeal to advanced users and teams looking to automate or visualize complex workflows directly within Gemini.
