Evidence mounts for Google to reveal a new Gemini model with agentic use case this week

Google appears to be preparing for another major AI-related announcement, with speculation centered around updates to its Gemini model and new features in its AI ecosystem. Over the weekend, users on lmarena noticed a model named "Nebula," which identifies itself as a Google creation and performs well on certain prompts. This has led to speculation that Nebula could be an updated version of Gemini 2.0 Pro, potentially incorporating advanced reasoning capabilities.

UPDATE: Some reports from Reddit suggest that Google started rolling out Gemini 2.5 Pro (experimental) with "Thinking" capabilities 🔥

Adding to the intrigue, Google employees have hinted at upcoming developments, such as Logan's cryptic tweet mentioning "Gemini," which aligns with past signals preceding launches.

Code analysis reveals references to a March 18 launch date, suggesting that Google may have planned an earlier release but delayed it. This raises the likelihood of announcements today or tomorrow. One confirmed change is the addition of a new model in the drop-down menu, consistent with previous launches. Additionally, UI updates hidden behind feature flags point to a new "Agents" toolbox on Gemini's web interface, featuring options like Deep Research and Canvas. While these tools are not entirely new, the "Agents" label hints at potential functionality expansions.

Two notable notices have also been added to the code: "AI Transparency Notice" and "Email Opt-in Discovery Card." The former likely reflects Google's commitment to transparency in AI development, while the latter suggests users may soon subscribe to email notifications for specific tasks or model completions. This feature could integrate with Deep Research or other tools.

Looking ahead, several possibilities emerge for this week's releases:

• Scheduled Prompts: A feature allowing users to schedule periodic prompt executions, similar to ChatGPT tasks.

• Creative Partner Tool: An image editing tool enabling inpainting capabilities that may expand Gemini's native image output from AI Studio to Gemini Web.
• Video Generation: Potential integration of VEO2 technology for animating images generated by Imagen3 models.

Based on the strings found in the code referencing unreleased features, there’s high confidence that at least some of these updates will be announced imminently:

• 🔥 New model launch (99% confidence): This points to a new model announcement—possibly Gemini 2.0 Pro Thinking (30% chance) or a new model that supports Scheduled Tasks (70% chance).
• AI transparency notice (90% confidence): Details remain unclear, but it links to the Google account chooser, suggesting it may relate to a future enterprise toggle allowing users to switch between personal and workspace accounts.
• Email opt-in (90% confidence): Likely tied to the Scheduled Prompts feature, enabling users to receive email notifications when tasks are completed.
• Video Generation (50% confidence): Indicates upcoming video generation features, potentially powered by VEO2 and possibly launching with limited availability.

Altogether, the strongest signals point to a new model launch, email opt-in for scheduled tasks, and potentially some level of a new toggle for enterprise users, with video generation still a possibility, though less certain.